framework,version,device,op_name,kernel_source,gemm_dtype,m,n,k,latency
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,12288,11.389999389648438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,16384,27.375322129991318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,10240,16.378399319118923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,16384,17.71809556749132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,12288,20.330529106987846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,10240,9.67170884874132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,8192,7.701436360677083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,8192,13.386394076877169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,7168,7.681842380099827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,7168,11.601270887586805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,6144,6.051374223497178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,6144,9.82015906439887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,5120,8.406802707248264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,5120,4.665592193603516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,16384,14.339836968315971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,4096,6.759063296847874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,4096,3.656731499565972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,12288,10.273924933539496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,10240,8.622781541612413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,8192,6.973679860432942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,3584,5.873883565266927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,3584,3.0943963792588978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,3072,2.5492070515950522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,3072,5.059245215521918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,7168,6.119177924262153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,6144,5.308908250596788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,2560,4.2445322672526045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,5120,4.44392818874783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,2560,2.1207093132866754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,2048,3.362751007080078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,4096,3.5103982289632163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,2048,1.6240737703111436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,1536,2.4879769219292536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,1536,1.2693333095974393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,1024,1.733039961920844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,1024,0.9206213421291776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,768,1.2598649130927193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,768,0.7707200050354004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,3584,3.1160426669650607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,512,0.9230168660481771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,512,0.5958053270975748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,2560,2.2288470798068576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,256,0.5199742317199707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,256,0.47499291102091473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,3072,2.7055119408501516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,128,0.4265608787536621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,128,0.4043911033206516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,64,0.3532417880164252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,64,0.42722755008273655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,65536,32,0.3471208943261041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,65536,32,0.4308577908409967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,1536,1.3805440266927083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,2048,1.8360640207926433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,65536,27.202348497178818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,65536,20.27570258246528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,512,0.5348533524407281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,16384,3.9460932413736978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,16384,6.731379614935981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,12288,5.02475823296441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,768,0.7153440051608615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,12288,2.946943071153429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,10240,4.221652560763888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,10240,2.4600577884250217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,1024,0.8884408738878039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,8192,3.3719974093967013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,8192,1.910777833726671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,256,0.3992311159769694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,7168,2.946608861287435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,7168,1.6957751380072699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,65536,128,0.3765422238243951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,6144,2.4510436587863498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,6144,1.426269319322374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,65536,14.834604899088541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,5120,2.078716490003798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,5120,1.184332423739963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,16384,3.508454216851128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,4096,1.6381137635972765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,4096,0.8481040000915527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,12288,2.656928804185655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,3584,1.427286254035102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,3584,0.7404738002353244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,10240,2.215846167670356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,3072,1.1659048928154838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,3072,0.586343977186415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,8192,1.8112017313639324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,2560,0.9683653513590494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,2560,0.49197864532470703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,7168,1.5308621724446614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,2048,0.7577342457241483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,2048,0.39413777987162274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,6144,1.2900106641981337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,1536,0.5812613169352213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,1536,0.3014204502105713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,5120,1.0705191294352214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,1024,0.3939990997314453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,1024,0.22638045416937935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,4096,0.8572862413194445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,768,0.3090195655822754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,3584,0.7116150856018066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,768,0.18943733639187285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,512,0.2205351061291165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,512,0.15332178274790445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,2560,0.5054871241251627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,256,0.1313777764638265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,256,0.12621155050065783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,3072,0.6144942177666558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,128,0.09313866827223037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,128,0.10912888579898411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,64,0.08716355429755317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,64,0.11518577734629314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,16384,32,0.0871608919567532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,16384,32,0.11524266666836208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,1536,0.32392533620198566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,2048,0.4228275616963704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,768,0.19206844435797799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,65536,20.14815436469184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,65536,15.417516072591146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,16384,4.917065090603299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,16384,2.994079165988498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,12288,3.6790317959255643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,12288,2.2061680687798395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,1024,0.23867644204033744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,10240,3.0643102857801647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,10240,1.8310489654541016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,256,0.09950844446818034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,8192,2.4200657738579645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,8192,1.3838941786024306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,512,0.1522293355729845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,7168,2.0919626024034286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,7168,1.263887087504069
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,16384,128,0.09236977497736613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,6144,1.7215742535061305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,6144,0.9920541975233289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,65536,11.038044399685331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,5120,1.4355715645684137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,5120,0.8551191223992242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,16384,2.687891642252604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,4096,1.1464355256822374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,4096,0.648746649424235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,12288,2.060455110337999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,3584,1.0097217559814453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,3584,0.5494533114963108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,10240,1.6342835956149633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,8192,1.2976889080471463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,3072,0.8223448859320747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,3072,0.45836443371242946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,2560,0.6956559816996256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,2560,0.37596978081597227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,7168,1.1421857410007052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,2048,0.5503964424133301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,2048,0.3003013398912218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,6144,0.8959031634860568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,5120,0.7721982002258301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,1536,0.4138791031307644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,1536,0.22881688012017143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,1024,0.28676266140407985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,1024,0.1724382241566976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,3584,0.5450222227308485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,768,0.22674844000074598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,768,0.14741422070397273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,4096,0.6210746765136719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,512,0.16579911443922254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,2560,0.39504088295830625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,512,0.12319289313422309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,256,0.10171111424763997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,256,0.09744177924262153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,3072,0.4590257538689508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,128,0.06884533166885376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,128,0.08477333519193862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,64,0.06270933151245117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,1536,0.24909777111477324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,768,0.14638045099046496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,64,0.08912711011038886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,12288,32,0.06328177452087402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,12288,32,0.0901671118206448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,2048,0.3201457659403483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,65536,13.481540256076388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,65536,17.233837551540798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,16384,4.136005401611328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,16384,2.521134270562066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,12288,3.0652258131239147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,12288,1.7555529276529949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,1024,0.18157154983944365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,10240,2.454266654120551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,10240,1.5231182310316298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,512,0.11557155185275608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,8192,2.043535020616319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,128,0.07049600283304851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,12288,256,0.07523288991716173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,8192,1.1879119873046875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,7168,1.7885386149088542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,7168,1.0457519955105252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,6144,1.4039920171101887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,6144,0.8657111061943902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,16384,2.259770711263021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,5120,1.1804702546861436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,65536,9.203088548448351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,5120,0.7083839840359158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,12288,1.6421715418497722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,4096,0.5333111021253798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,4096,0.9854364395141602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,3584,0.8254186842176648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,3584,0.45306046803792316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,10240,1.3515262603759766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,3072,0.705352889166938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,3072,0.38369332419501406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,8192,1.0174212985568576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,2560,0.5682995584275987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,2560,0.31389689445495605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,7168,0.9278924730088977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,2048,0.47112443712022567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,5120,0.63064087761773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,2048,0.25760889053344727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,6144,0.7988142437405057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,1536,0.3580177889929877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,1536,0.19799999396006265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,1024,0.24724978870815703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,1024,0.14822044637468126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,3584,0.45161067114935977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,768,0.19214488400353325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,768,0.12576266129811606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,2560,0.33015110757615834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,4096,0.5185768869188097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,512,0.13871199554867214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,512,0.10381066799163818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,3072,0.3809191120995416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,256,0.08635733524958293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,256,0.08207911252975464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,2048,0.26602223184373647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,128,0.06180177794562446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,128,0.072781331009335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,64,0.05654578076468574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,64,0.07699111435148451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,768,0.12448622120751275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,1536,0.20980000495910645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,10240,32,0.05696088737911648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,10240,32,0.07696533203125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,65536,7.864315456814236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,65536,13.250189039442274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,16384,3.3425680796305337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,16384,2.0001129574245877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,12288,2.500440809461806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,12288,1.4377164840698242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,1024,0.15601600541008845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,10240,2.0672825707329645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,10240,1.1699342727661133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,512,0.09854577647315131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,8192,1.6429973178439672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,8192,0.9375502268473307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,128,0.06118488974041409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,10240,256,0.06487822532653809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,7168,0.8212186495463053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,7168,1.4294284184773762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,6144,1.2315919664171007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,6144,0.6739884482489692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,16384,1.8054080539279516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,65536,7.408827039930556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,5120,0.9648097356160482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,5120,0.5656542248196071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,4096,0.762773354848226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,4096,0.42836088604397243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,12288,1.249243524339464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,3584,0.6396737628512913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,3584,0.355039119720459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,10240,1.0463662677341037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,3072,0.560132450527615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,3072,0.30528799692789715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,8192,0.8296684688991971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,2560,0.4770035743713379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,2560,0.2582284344567193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,7168,0.7210559844970703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,2048,0.3767128785451253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,2048,0.20393244425455728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,6144,0.6127404636806911
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,1536,0.28594311078389484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,1536,0.15685955683390299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,5120,0.518164422776964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,1024,0.19779911306169298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,1024,0.11711822615729438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,3584,0.3702897760603163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,768,0.15337244669596353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,768,0.1005600028567844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,2560,0.2682195504506429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,512,0.10966933435863918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,512,0.08037244611316256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,4096,0.4143253432379828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,256,0.0661404463979933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,256,0.06898133622275458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,3072,0.31063734160529244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,128,0.04971999923388163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,128,0.05966222286224365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,64,0.048727999130884804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,64,0.06122933493720161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,8192,32,0.04618311259481642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,8192,32,0.06062044699986776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,2048,0.22338578436109754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,65536,11.602220323350695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,65536,7.009507497151692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,1536,0.17252088917626274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,16384,2.851897769504123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,16384,1.7556080288357203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,768,0.10236889123916626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,12288,2.1409369574652777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,12288,1.2329271104600694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,1024,0.12575556172264948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,10240,1.658823119269477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,10240,1.01837158203125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,128,0.05004177822007073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,8192,1.3785342110527887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,256,0.05270222160551283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,8192,0.8045973247951932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,7168,1.1346284018622506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,7168,0.686830202738444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,8192,512,0.07987288633982341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,6144,0.9609759648640951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,6144,0.6037715805901421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,65536,6.576493157280816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,5120,0.7968311309814453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,5120,0.47565333048502606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,16384,1.542001830206977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,4096,0.6407564481099447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,4096,0.37939021322462296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,12288,1.0811298158433702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,3584,0.5598399904039171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,3584,0.32209600342644584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,10240,0.9168844223022461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,3072,0.46803469128078884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,3072,0.27550578117370605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,8192,0.761756420135498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,2560,0.39622044563293457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,2560,0.2274657885233561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,7168,0.6420799891153971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,2048,0.31833600997924805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,2048,0.18252710501352945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,6144,0.5389368798997667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,1536,0.24158133400811088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,1536,0.14017511738671198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,5120,0.4670871098836263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,1024,0.17233422067430285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,1024,0.10636088583204482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,4096,0.3651422129737006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,768,0.1318355533811781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,768,0.09015555514229669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,3584,0.33185866143968373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,512,0.09861510992050171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,512,0.0753377808464898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,2560,0.2398515542348226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,256,0.05899910794364082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,256,0.06089155541525947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,3072,0.2825128767225477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,128,0.04486844605869717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,128,0.055086221959855824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,64,0.040307554933759905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,64,0.05758400095833672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,7168,32,0.04114755656984117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,7168,32,0.05600444475809733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,1536,0.15248889393276638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,65536,9.997749328613281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,768,0.09012711048126221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,65536,6.040091620551215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,2048,0.19649688402811685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,16384,1.4148231082492406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,16384,2.4734168582492404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,12288,1.8017039828830297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,12288,1.092224015129937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,1024,0.11162755224439834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,10240,1.483936839633518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,10240,0.8593413564893934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,256,0.047741333643595375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,8192,1.2085520426432292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,8192,0.6986675792270236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,512,0.07056711117426555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,7168,0.964721785651313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,7168,0.5876764191521539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,7168,128,0.04426666763093737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,6144,0.8821226755777994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,6144,0.5086702240837945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,65536,5.704490661621094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,5120,0.6997360123528374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,5120,0.4094453387790256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,16384,1.3598106172349718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,4096,0.547776910993788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,4096,0.3210337691836887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,12288,1.0043777889675565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,3584,0.47649865680270725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,3584,0.27785155508253306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,10240,0.799507564968533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,3072,0.4073973231845432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,3072,0.23719912105136445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,8192,0.6189742088317871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,2560,0.33590133984883624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,2560,0.19461333751678467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,7168,0.5677013397216797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,2048,0.27955733405219185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,2048,0.1616426706314087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,6144,0.4653049045138889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,1536,0.20700444115532768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,1536,0.12237955464257134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,5120,0.39593156178792316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,1024,0.14637777540418836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,1024,0.09234044286939834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,3584,0.28576355510287815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,4096,0.32074043485853404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,768,0.11662933561537002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,768,0.07885866694980197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,512,0.08413955238130356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,512,0.06365688641866048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,2560,0.20708089404635957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,256,0.0510791109667884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,256,0.05247111121813456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,3072,0.24354399575127494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,128,0.03934489025010003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,128,0.04668000009324816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,64,0.03488622109095255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,64,0.04890844557020399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,6144,32,0.03577155537075467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,6144,32,0.04807733496030172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,1536,0.13467644320593938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,65536,8.266136169433594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,65536,5.040112813313802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,2048,0.16788089275360107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,768,0.07881244685914782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,16384,2.0316880544026694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,16384,1.1852462556627061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,12288,1.4183520211113825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,12288,0.8984195921156141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,1024,0.09697155819998847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,10240,1.2144080268012154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,10240,0.7241848839653863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,128,0.038464887274636164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,8192,0.8942569096883138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,256,0.0414737794134352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,8192,0.5764426655239528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,7168,0.8328773180643717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,7168,0.5007333225674099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,6144,512,0.06198666493097941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,6144,0.7055155436197916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,6144,0.42214134004380965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,65536,4.8372578091091585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,5120,0.5805493460761176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,5120,0.3374062114291721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,16384,1.0952471627129448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,4096,0.4652986526489258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,4096,0.2727128929562039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,12288,0.7908933427598742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,3584,0.3961742189195421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,3584,0.2354524400499132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,10240,0.6937724219428169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,3072,0.346327993604872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,3072,0.20288888613382974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,8192,0.533050643073188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,2560,0.28668710920545787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,2560,0.16675378216637504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,6144,0.40181777212354874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,2048,0.23374311129252115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,2048,0.1363662216398451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,7168,0.47211111916436094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,1536,0.1790133317311605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,1536,0.10659466849433051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,5120,0.342940436469184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,1024,0.12613688574896917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,1024,0.07950577470991346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,3584,0.246434662077162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,768,0.0997919970088535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,768,0.06883022520277235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,4096,0.2746515538957384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,512,0.07195111115773518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,512,0.05652533637152778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,2560,0.17840533786349824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,256,0.04612266686227587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,256,0.04622755448023478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,3072,0.20639555984073213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,128,0.03361777795685662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,128,0.041800889703962535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,64,0.02990844514634874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,64,0.043452445003721446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,5120,32,0.031063109636306763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,5120,32,0.043014221721225314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,2048,0.1449848943286472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,65536,6.457628462049697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,65536,4.124595642089844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,1536,0.11449778079986572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,16384,1.5109902487860785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,768,0.06764444377687243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,16384,0.9819439782036675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,12288,1.0921849144829643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,12288,0.6990559895833334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,1024,0.08398489157358806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,10240,0.9564986758761935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,10240,0.5809662077162001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,256,0.03626133335961236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,8192,0.7400568856133355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,8192,0.4607315593295627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,128,0.0336168905099233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,7168,0.6354275809393989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,7168,0.3949208789401584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,5120,512,0.0535093347231547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,6144,0.5449875725640191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,6144,0.33677511745029026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,65536,3.925132327609592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,5120,0.44262311193678117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,5120,0.27245955997043186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,16384,0.9154080284966363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,4096,0.35498666763305664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,4096,0.22005955378214517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,12288,0.6629021962483724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,3584,0.32123110029432506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,3584,0.19453599717881945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,10240,0.5466168721516927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,3072,0.2721128993564182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,3072,0.1674640046225654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,8192,0.4424560070037842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,2560,0.2289795610639784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,2560,0.13768800099690756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,7168,0.3878382311926947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,2048,0.18499378363291422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,2048,0.11109689209196304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,6144,0.33161600430806476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,1536,0.1416542265150282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,1536,0.0872560011015998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,5120,0.28271910879347056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,1024,0.10128621922598945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,1024,0.06584711207283868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,4096,0.222326225704617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,3584,0.19905243979560006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,768,0.0776853362719218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,768,0.05666489071316189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,512,0.056806219948662647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,512,0.04565777712398105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,2560,0.14755377504560682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,256,0.0362115568584866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,256,0.03882666760020786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,3072,0.1736826631757948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,128,0.028024001253975764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,128,0.03438933359252082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,64,0.024454222785101995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,64,0.03522488805982802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,4096,32,0.025359110699759588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,4096,32,0.03496177660094367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,2048,0.12089688248104519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,65536,5.721557193332249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,65536,3.622528923882378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,1536,0.09509066740671794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,16384,1.2337208853827584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,16384,0.8318950865003797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,768,0.057017776701185435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,12288,1.0040150748358834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,12288,0.6417199770609537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,1024,0.06971466541290283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,10240,0.801977793375651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,10240,0.5293048752678765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,256,0.030642665094799463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,128,0.028485334581798975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,8192,0.6165946854485406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,8192,0.40610400835673016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,7168,0.5502479871114095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,7168,0.3503759966956244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,4096,512,0.04430400000678169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,6144,0.4639520115322537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,6144,0.2962097856733534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,65536,3.5929429796006946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,5120,0.39974045753479004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,5120,0.24927287631564668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,16384,0.8165342013041178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,4096,0.3185848924848768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,4096,0.2014231152004666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,12288,0.6047333081563314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,3584,0.27979911698235405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,3584,0.1763413349787394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,10240,0.4997599919637044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,3072,0.24397065904405382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,3072,0.1537582212024265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,8192,0.39600621329413516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,2560,0.20037244425879586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,2560,0.12637066841125488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,6144,0.2988319926791721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,2048,0.16222933928171793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,2048,0.10279733604855007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,7168,0.34480622079637313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,1536,0.12279200553894043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,1536,0.07941955327987671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,5120,0.24744534492492676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,1024,0.08670488993326823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,1024,0.058957331710391574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,4096,0.2052711115943061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,768,0.06996622350480822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,768,0.05236266718970405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,3584,0.18052800496419272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,512,0.05113866594102648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,512,0.04232355621125963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,2560,0.13489333788553873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,256,0.03264355659484863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,256,0.035623110002941556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,3072,0.1558613379796346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,128,0.025366221865018208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,128,0.03207555413246155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,64,0.02309333284695943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,64,0.03255733185344272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3584,32,0.02402488887310028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3584,32,0.03211733367707994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,1536,0.08664000034332275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,65536,4.859844631618924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,65536,3.156992806328667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,768,0.0506879985332489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,16384,1.07495911916097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,16384,0.7316222190856934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,2048,0.10937244362301296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,12288,0.8284079763624402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,12288,0.5528711213005914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,1024,0.06276355849372016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,10240,0.6805760065714518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,10240,0.4422924518585205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,256,0.027112889620992873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,8192,0.5369324684143066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,128,0.024844444460339014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,8192,0.35670489735073513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,7168,0.470811578962538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3584,512,0.039213332864973284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,7168,0.3087795575459798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,6144,0.405872901280721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,6144,0.262792878680759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,65536,3.1044960021972656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,5120,0.3425262239244249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,5120,0.21914133760664198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,16384,0.6887475649515787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,4096,0.2729102240668403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,4096,0.17896533012390137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,12288,0.5145466592576768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,3584,0.23654490047030977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,3584,0.15772799650828043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,10240,0.4343253241644965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,3072,0.2057795524597168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,3072,0.13635467158423528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,8192,0.34638222058614093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,2560,0.16985778013865152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,2560,0.11176178190443252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,7168,0.31149421797858345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,2048,0.14222399393717447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,2048,0.09127644697825114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,6144,0.26591287718878853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,1536,0.10904533333248562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,1536,0.07059378094143338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,5120,0.22255023320515951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,1024,0.0763066675927904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,1024,0.051961776283052236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,3584,0.1600266695022583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,768,0.060119999779595264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,768,0.04676533407635159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,4096,0.17981421947479248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,512,0.044441776143180005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,512,0.037258668078316584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,2560,0.11773244539896648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,256,0.029136889510684546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,256,0.031168888012568157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,3072,0.13651288880242243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,128,0.02163377735349867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,128,0.02828622195455763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,64,0.020255110330051847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,64,0.02853155467245314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,3072,32,0.020938666330443487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,3072,32,0.028283556302388508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,2048,0.09711288743548924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,65536,4.039625803629558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,65536,2.6991102430555554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,768,0.044300443596310086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,16384,0.938818613688151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,1536,0.07687911060121325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,16384,0.6338559786478678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,12288,0.6833368937174479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,12288,0.47041066487630206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,1024,0.05576444334454007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,10240,0.5579279793633355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,10240,0.38810133934020996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,128,0.021298666795094807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,256,0.024247111545668706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,8192,0.4442231125301785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,8192,0.30754489368862575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,3072,512,0.03457955519358317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,7168,0.39018845558166504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,7168,0.26703911357455784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,6144,0.33414755927191836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,6144,0.22870932685004342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,65536,2.6945243411593967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,5120,0.2790986696879069
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,5120,0.1919271151224772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,16384,0.6032337612575954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,4096,0.2237022187974718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,4096,0.15692355897691515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,12288,0.4574044545491536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,3584,0.20324088467491996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,3584,0.13860800531175402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,10240,0.37787289089626735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,3072,0.1734213299221463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,3072,0.11943999926249187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,8192,0.30649156040615505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,2560,0.14684266514248318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,2560,0.09904533624649048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,7168,0.2657431231604682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,2048,0.11787022484673394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,2048,0.0804542236857944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,6144,0.22919556829664442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,1536,0.09121777613957723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,1536,0.0622346666124132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,5120,0.19359200530582002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,1024,0.06474666463004218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,1024,0.04604711135228475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,4096,0.1547057761086358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,768,0.051029334465662636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,768,0.041514668199751116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,3584,0.1389297776752048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,512,0.03872444563441806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,512,0.03295288814438714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,2560,0.10301333003573948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,256,0.02643288837538825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,256,0.027635556128289964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,3072,0.12117866675059001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,128,0.019713777634832595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,128,0.02504088812404209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,64,0.016120000018013846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,64,0.025283555189768474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2560,32,0.016851555969980028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2560,32,0.025174222058720056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,768,0.03925155599912008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,2048,0.08548711405860053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,65536,3.2105155520968967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,65536,2.2739573584662542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,16384,0.7066533300611707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,16384,0.5294168790181478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,1536,0.06677422258589003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,12288,0.5275777710808648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,12288,0.39256710476345485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,1024,0.04893955588340759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,10240,0.44059377246432835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,10240,0.32592444949679905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,256,0.021551999780866835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,8192,0.34760355949401855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,8192,0.2596000035603841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,512,0.03085600005255805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,7168,0.30647288428412545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,7168,0.22808443175421822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2560,128,0.018639999959203932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,6144,0.2678479883405897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,6144,0.19422311253017852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,65536,2.2303199768066406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,5120,0.2308879958258735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,5120,0.16396978166368273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,16384,0.512330691019694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,4096,0.18005155192481148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,4096,0.13410754998524985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,12288,0.3821573257446289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,3584,0.1609813372294108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,3584,0.1182568868001302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,10240,0.31719199816385907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,3072,0.13718667295244005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,3072,0.10201688607533772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,8192,0.2557964324951172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,2560,0.11583022276560466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,2560,0.08302044206195407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,7168,0.22499910990397134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,2048,0.09330755472183228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,2048,0.06708977619806926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,6144,0.19162933031717935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,1536,0.07210310962465075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,1536,0.052516443861855396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,5120,0.16317421860165066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,1024,0.05223200056287977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,1024,0.039035555389192365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,3584,0.11861422326829697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,768,0.04164888792567783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,768,0.035587555832333036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,4096,0.13272089428371853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,512,0.03135022189882066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,512,0.02749155627356635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,2560,0.08740444315804376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,256,0.021545777718226116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,256,0.02313866714636485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,3072,0.10284711254967584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,128,0.016531555189026725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,128,0.020780444145202637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,64,0.013595555391576556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,64,0.021439111895031396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,2048,32,0.013795555465751223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,2048,32,0.021387555532985266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,2048,0.07251644134521484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,1536,0.05745333433151245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,65536,2.3524320390489364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,65536,1.8032578362358942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,16384,0.5160382058885362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,16384,0.4414995511372884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,768,0.033570667107899986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,12288,0.396977769003974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,12288,0.32555288738674587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,1024,0.04193955659866333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,10240,0.32608622974819607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,10240,0.2733244366115994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,256,0.018358222312397428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,8192,0.2701102097829183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,8192,0.22169600592719185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,128,0.01571466690964169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,7168,0.23344977696736655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,7168,0.19214133421579996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,2048,512,0.026206221845414903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,6144,0.20050311088562012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,6144,0.16411466068691677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,65536,1.7941021389431422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,5120,0.1724755499098036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,5120,0.13908621999952528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,16384,0.4147973325517442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,4096,0.13702132966783312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,4096,0.11298755804697673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,12288,0.31187910503811306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,3584,0.1188711060418023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,10240,0.2616355684068468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,3584,0.10176622205310398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,3072,0.10301333003573948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,3072,0.08769333362579346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,8192,0.20982666810353598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,2560,0.08688444561428493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,2560,0.07165510786904229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,7168,0.1870115598042806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,2048,0.07304622067345513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,2048,0.05766222212049696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,5120,0.13694577746921116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,1536,0.05606400304370456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,1536,0.0446684459845225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,6144,0.16067643960316977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,1024,0.040399110979504056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,1024,0.03314044409328037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,4096,0.11188356081644694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,768,0.03261244297027588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,768,0.03048266636000739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,3584,0.09842399756113689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,512,0.024677332904603746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,512,0.02289688918325636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,2560,0.07425244649251302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,256,0.017688888642523024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,256,0.019132445255915325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,3072,0.08581866820653279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,128,0.013611555927329592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,128,0.017507554756270517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,64,0.01127377814716763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,64,0.017280000779363845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1536,32,0.011725333001878528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1536,32,0.017700443665186565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,2048,0.061095111899905734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,65536,1.5998559527926977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,65536,1.4492382473415797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,768,0.028671112325456407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,16384,0.3497022257910834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,16384,0.35326221254136825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,1536,0.048968887991375394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,12288,0.2678408887651232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,12288,0.263317346572876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,1024,0.03560444381501939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,10240,0.22000443935394287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,10240,0.2200791173511081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,128,0.013816888961527081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,8192,0.17437156041463217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,8192,0.18084800243377686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,256,0.016823111308945548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,7168,0.1559573279486762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,7168,0.15806489520602757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1536,512,0.023039110832744177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,6144,0.13596622149149576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,6144,0.13335822688208684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,65536,1.342440923055013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,5120,0.11365332868364121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,5120,0.11282222800784641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,16384,0.32208887736002606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,4096,0.09012177917692397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,4096,0.09410666757159764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,12288,0.24467200703091094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,3584,0.08046933015187581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,3584,0.08321422338485718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,10240,0.20609422524770102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,3072,0.07096799876954821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,3072,0.07200088765886095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,8192,0.16708089245690239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,2560,0.058740443653530545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,2560,0.0587902201546563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,7168,0.14663999610477024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,2048,0.047629333204693265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,2048,0.04691466689109802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,6144,0.1265217728084988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,1536,0.03749688797526889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,1536,0.03637155559327867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,5120,0.10811466640896267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,1024,0.026642666922675237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,1024,0.026526222626368206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,4096,0.08788355853822495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,3584,0.0778942240609063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,768,0.021448888712459143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,768,0.02557955516709222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,512,0.016397333807415433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,512,0.018800000349680584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,2560,0.05867822302712334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,256,0.011594666375054253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,256,0.01589866644806332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,3072,0.06829155815972222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,128,0.008551999926567078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,128,0.013832000394662222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,64,0.007344000041484833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,64,0.013939556148317126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,1024,32,0.007669332954618666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,1024,32,0.014102222190962898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,2048,0.049231112003326416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,65536,1.1573475731743705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,65536,1.2359049055311415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,1536,0.039000888665517174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,16384,0.2703297668033176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,768,0.02349066734313965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,16384,0.3049662113189697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,12288,0.20354221926795113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,12288,0.22785777515835234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,1024,0.028299556838141546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,10240,0.17316710948944092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,10240,0.19300533665551078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,256,0.014084445105658637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,8192,0.13741689258151585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,8192,0.15906843874189588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,128,0.011906666888131036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,7168,0.124309327867296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,7168,0.13600977261861166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,1024,512,0.018904000520706177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,6144,0.10617066754235162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,6144,0.11647644307878281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,65536,1.11407470703125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,5120,0.08956888649198745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,5120,0.09891733196046616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,16384,0.27883466084798175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,4096,0.07129155264960395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,4096,0.08229600058661567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,12288,0.21118222342597115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,3584,0.0640711122088962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,3584,0.073908441596561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,10240,0.17645155058966744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,3072,0.05387111173735725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,3072,0.06374044550789727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,8192,0.1425644424226549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,2560,0.04634399877654182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,2560,0.05271555648909675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,7168,0.12537777423858643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,2048,0.037392887804243304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,2048,0.0418888893392351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,6144,0.10901688867145115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,1536,0.029639111624823675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,1536,0.03204800022972955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,5120,0.09222666422526042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,1024,0.020980444219377305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,1024,0.023465777436892193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,3584,0.06786933210160997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,768,0.017302221722073026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,768,0.02292355563905504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,4096,0.07548355393939547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,512,0.013366222381591797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,512,0.016703110602166917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,2560,0.05109777715471056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,256,0.009094222552246517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,256,0.013832888669437833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,3072,0.059243553214603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,128,0.007167110840479533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,128,0.011967110964987012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,64,0.006172444257471297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,64,0.012316444681750404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,768,32,0.006502222269773483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,768,32,0.012498666842778524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,2048,0.042340444193945996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,65536,0.716733349694146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,65536,1.0469350814819336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,768,0.021182222498787776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,16384,0.18213510513305664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,16384,0.26289155748155385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,1536,0.033901333808898926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,12288,0.1351742214626736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,12288,0.19761244455973306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,1024,0.025016888976097107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,10240,0.11693778302934434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,10240,0.16213244862026638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,512,0.017303110824690927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,8192,0.09217511283026801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,8192,0.1334951056374444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,256,0.01312355531586541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,7168,0.0817377765973409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,7168,0.11656444602542454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,768,128,0.010628444453080496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,6144,0.06899911165237427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,6144,0.09995377726025051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,65536,0.9502301745944552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,5120,0.060702224572499595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,5120,0.08567111359702216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,16384,0.23637244436475965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,4096,0.048130667871899076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,4096,0.07143021954430474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,12288,0.17882488833533394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,3584,0.04271644353866577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,3584,0.06392621994018555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,10240,0.15001955297258165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,3072,0.036488887336519026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,3072,0.05680711401833428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,8192,0.12183554967244466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,2560,0.03184977836079068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,2560,0.04542577928966946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,7168,0.10793422328101264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,2048,0.02585244509908888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,2048,0.03527911172972785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,6144,0.09325422181023492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,1536,0.02033244404527876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,1536,0.02680533296532101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,5120,0.07932800054550171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,1024,0.015288889408111572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,1024,0.020051555501090158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,4096,0.06536355283525255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,768,0.012818666795889536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,768,0.02035466664367252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,3584,0.05810222360822889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,512,0.01016711112525728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,512,0.014037332601017423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,2560,0.0439226660463545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,256,0.007153777612580194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,256,0.012039999994966718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,3072,0.051113777690463595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,128,0.005963555640644497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,128,0.01052977806992001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,64,0.0053031109273433685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,64,0.010798222488827176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,512,32,0.005535111245181825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,512,32,0.011008888483047485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,2048,0.03654133280118307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,65536,0.4107057783338759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,65536,0.8687049018012153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,1536,0.02927910950448778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,16384,0.1032480001449585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,16384,0.2157671054204305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,768,0.01846222248342302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,12288,0.07689511113696627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,12288,0.16299466292063394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,1024,0.021870222356584337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,10240,0.06558399730258517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,10240,0.137609773212009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,128,0.009695111049546136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,8192,0.051720887422561646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,8192,0.11335910691155328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,256,0.011751111182901593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,7168,0.04622488882806566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,7168,0.09977689054277207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,512,512,0.015112888481881885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,6144,0.0403128895494673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,6144,0.08607288863923813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,65536,0.7890444331698947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,5120,0.03471466567781236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,5120,0.07362222009234957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,16384,0.20484533574846056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,4096,0.02785244584083557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,4096,0.06105955441792806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,12288,0.15268622504340276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,3584,0.024273778001467388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,3584,0.05447466505898369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,10240,0.12833778063456217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,3072,0.021087111698256597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,3072,0.04734222094217936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,8192,0.10387111372417873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,2560,0.01832444469134013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,2560,0.03784888982772827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,7168,0.09166400300131904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,2048,0.015103111664454142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,2048,0.029847111966874864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,6144,0.07906133598751493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,1536,0.012249777714411417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,1536,0.022283555732833013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,5120,0.06707644462585449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,1024,0.009740444521109263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,1024,0.017158221867349412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,4096,0.054392887486351855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,768,0.008452444440788692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,768,0.01759466694460975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,3584,0.048898667097091675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,512,0.007137777904669444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,512,0.012072000238630505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,2560,0.036601778533723615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,256,0.004993777722120285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,256,0.010254222485754225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,3072,0.0427724454138014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,128,0.004382222063011593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,128,0.0090515555606948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,64,0.003927111211750242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,64,0.009275555610656738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,256,32,0.004292444636424382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,256,32,0.009480888644854227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,2048,0.03030666708946228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,65536,0.2849137783050537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,65536,0.8309964603847928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,1536,0.024384000235133704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,16384,0.07832533121109009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,16384,0.20937332842085096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,768,0.015588444140222339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,12288,0.054224888483683266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,12288,0.15911111566755506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,1024,0.01847466660870446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,10240,0.04602933261129591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,10240,0.13484533627827963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,256,0.010084444450007545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,8192,0.03793333305252923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,8192,0.10976177453994751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,512,0.012989333106411828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,7168,0.033253331979115806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,7168,0.09736710786819458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,256,128,0.007124444676770105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,6144,0.029005332125557795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,6144,0.08427644438213772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,65536,0.7119777997334799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,5120,0.02754844393995073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,5120,0.07183644506666395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,16384,0.18482844034830728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,4096,0.021718222233984206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,4096,0.060455110337999135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,12288,0.13909955819447836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,3584,0.018724444839689467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,3584,0.05342310998174879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,10240,0.11756800280676948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,3072,0.016728000508414376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,3072,0.046500444412231445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,8192,0.09571644332673813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,2560,0.014459555347760519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,2560,0.03667733404371474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,7168,0.08428444465001424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,2048,0.011928000383906893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,6144,0.07287200291951497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,2048,0.028279112444983587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,1536,0.009861333502663506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,1536,0.02200177808602651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,5120,0.061588446299235024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,1024,0.00775911079512702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,1024,0.016865778300497267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,4096,0.050223999553256564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,768,0.006833777659469181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,768,0.017478222648302715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,3584,0.04511377877659268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,512,0.0058373332851462895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,512,0.012054222325483957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,2560,0.033561779393090144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,3072,0.039481778939565025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,256,0.004080888711743885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,256,0.009619555539555019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,128,0.00360977773865064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,128,0.008801778157552084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,64,0.0033244445092148254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,64,0.009004444711738164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,128,32,0.0036728887094391715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,128,32,0.009015111459626092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,65536,0.2683386537763807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,65536,0.8618239826626248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,16384,0.07940266529719035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,16384,0.21516266134050158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,12288,0.051710221502516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,12288,0.1617679993311564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,10240,0.044403556320402354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,10240,0.13655466503567165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,8192,0.03622666663593716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,8192,0.11181422074635823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,7168,0.031926221317715116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,7168,0.09878489043977524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,6144,0.028027554353078205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,6144,0.084824886586931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,5120,0.024287111229366724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,5120,0.07254933648639254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,4096,0.020454222957293194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,4096,0.06069066789415148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,3584,0.018026666508780587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,3584,0.05442666676309374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,3072,0.016044444508022733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,3072,0.04734755555788676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,2560,0.013979555832015144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,2560,0.03805511196454366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,2048,0.01200355589389801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,2048,0.029351999362309773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,1536,0.009859555297427708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,1536,0.022669333550665114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,1024,0.007671111159854465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,1024,0.017273777061038546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,768,0.00665155549844106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,768,0.01735822194152408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,512,0.00546577779783143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,512,0.011881777809725868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,256,0.0037439999481042228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,256,0.00976533359951443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,128,0.0033857776886887024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,128,0.008859555754396651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,64,0.003237333355678452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,64,0.008947555389669206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,64,32,0.0032266666077905228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,64,32,0.008960000342792934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,65536,0.2612817817264133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,65536,0.8698026869032117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,16384,0.07711644305123223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,16384,0.21936710675557455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,12288,0.05408622158898247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,12288,0.16496622562408447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,10240,0.046181334389580615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,10240,0.13806488778856066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,8192,0.03501333461867438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,8192,0.11296978261735703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,7168,0.03086755673090617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,7168,0.10042666726642185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,6144,0.027133334014150832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,6144,0.08549155791600545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,5120,0.023576888773176406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,5120,0.07469599776797824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,4096,0.019780443774329293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,4096,0.061583108372158475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,3584,0.017535110314687092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,3584,0.0546284450425042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,3072,0.01551822159025404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,3072,0.04706222150060865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,2560,0.013609777722093793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,2560,0.03871644536654154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,2048,0.011360888679822287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,2048,0.02933244572745429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,1536,0.009359999663299983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,1536,0.022854222191704646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,1024,0.007261333366235097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,1024,0.01737155516942342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,768,0.006362666686375936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,768,0.017501332693629794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,512,0.0052168890833854675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,512,0.011883555187119378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,256,0.00360977773865064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,256,0.009873777627944946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,128,0.0032417778339650896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,128,0.008936888641781276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,64,0.0030462222380770575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,64,0.008961777720186446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8192,32,32,0.003116444374124209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8192,32,32,0.008929777476522658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,2048,0.027615999182065327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,16384,6.8842578464084205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,16384,13.688681708441841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,16384,6.991373274061416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,12288,5.2370177374945746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,12288,9.846120198567709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,12288,5.401390923394097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,10240,4.35972637600369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,10240,8.186133490668402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,10240,4.443838331434462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,8192,3.501138687133789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,8192,6.56040276421441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,8192,3.4952320522732205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,7168,3.0822035471598306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,7168,3.0163449181450734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,7168,5.763502332899306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,6144,2.5874337090386286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,6144,2.5085555182562933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,6144,4.949145846896701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,5120,2.1897705925835504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,5120,4.146144866943359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,5120,2.0339688195122614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,4096,3.329805374145508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,4096,1.7264080047607422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,4096,1.599130630493164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,3584,2.9282879299587674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,3584,1.4878764682345922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,3584,1.3868746227688258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,3072,1.1907421747843425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,3072,2.4864383273654513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,3072,1.176545778910319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,2560,2.050632052951389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,2560,1.0585307015313041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,2560,0.9251182344224719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,2048,0.8351493411593967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,2048,1.5695679982503254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,2048,0.7613466580708822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,1536,1.1604142718844943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,1536,0.6324257850646973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,1536,0.5842889149983724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,1024,0.7590790854560004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,1024,0.4394071102142334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,1024,0.43479379018147785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,768,0.6088159879048666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,768,0.3595048851437039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,768,0.33832089106241864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,512,0.2506222195095486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,512,0.43264267179701066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,512,0.29270844989352757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,256,0.26271022690667045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,256,0.24151645766364205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,256,0.19900888866848418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,128,0.19645867082807753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,128,0.22017155753241646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,65536,128,0.19282044304741752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,64,0.1790106693903605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,64,0.22943909962972006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,65536,32,0.17722311284806994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,65536,32,0.23425687683953178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,65536,7.130116780598958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,65536,13.954935709635416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,65536,7.2523710462782125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,16384,1.6977440516153972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,16384,3.3607076009114585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,16384,1.8239146338568792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,12288,2.4937947591145835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,12288,1.2688364452785916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,12288,1.2632480197482638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,10240,2.067985746595595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,10240,1.0759315490722656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,10240,1.0419493781195748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,8192,1.6443022622002497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,8192,0.8288133409288195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,8192,0.8204008738199869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,7168,0.677818669213189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,7168,0.7167386478847928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,7168,1.374496036105686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,6144,1.205134179857042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,6144,0.5985937648349339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,6144,0.5935333569844564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,5120,0.9465750588311089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,5120,0.4927511215209961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,5120,0.48275020387437606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,4096,0.39472089873419863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,4096,0.7783270941840278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,4096,0.3780337704552545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,3584,0.3544062243567572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,3584,0.643936898973253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,3584,0.33560978041754824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,3072,0.5740008884006077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,3072,0.28704622056749135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,3072,0.29947021272447377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,2560,0.2588711049821642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,2560,0.46944088406032985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,2560,0.24137155214945474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,2048,0.3798862298329671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,2048,0.20822755495707193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,2048,0.19665688938564727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,1536,0.2882942093743218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,1536,0.1621662245856391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,1536,0.1538568867577447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,1024,0.20024889045291475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,1024,0.11695999569363064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,1024,0.11650044388241237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,768,0.15266933706071642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,768,0.09248799747890896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,768,0.0979306697845459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,512,0.0692231125301785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,512,0.11030044158299764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,512,0.08013066980573866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,256,0.049504889382256397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,256,0.0662097798453437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,256,0.06667200062010023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,128,0.04935288760397169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,16384,128,0.04900000161594815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,128,0.06113333172268338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,64,0.0455991095966763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,64,0.06329066885842217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,16384,32,0.04650488826963636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,16384,32,0.06419199705123901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,65536,5.3925281100802955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,65536,9.738825480143229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,65536,5.519441816541884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,16384,1.2930089102851021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,16384,2.343247095743815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,16384,1.3366453382703993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,12288,1.7658987045288086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,12288,0.9530870649549695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,12288,0.9756986829969617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,10240,1.4075057771470811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,10240,0.745939572652181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,10240,0.7942738003200955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,8192,1.1704870859781902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,8192,0.5884773466322157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,8192,0.6345280011494955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,7168,0.9582079781426324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,7168,0.5202044381035699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,7168,0.5358399815029568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,6144,0.44552887810601133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,6144,0.44340888659159344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,6144,0.8053199979994031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,5120,0.3718782265981038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,5120,0.6929564476013184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,5120,0.3692399925655789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,4096,0.5340044233534071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,4096,0.2933742205301921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,4096,0.3056151072184245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,3584,0.26579377386305064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,3584,0.47672446568806964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,3584,0.2644675572713216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,3072,0.40943288803100586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,3072,0.23232356707255045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,3072,0.2233182324303521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,2560,0.19785332679748535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,2560,0.33653778500027126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,2560,0.186553782886929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,2048,0.2764693366156684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,2048,0.15175644556681314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,2048,0.159551117155287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,1536,0.12588977813720703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,1536,0.20800799793667266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,1536,0.11934399604797363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,1024,0.09007466501659817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,1024,0.14719200134277344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,1024,0.09081333213382298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,768,0.07260977559619479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,768,0.11398488945431179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,768,0.07678400145636664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,512,0.05460888809627957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,512,0.08386400010850693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,512,0.06522044208314684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,256,0.05104177859094408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,256,0.0391866664091746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,256,0.052935110198126904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,128,0.03922577699025472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,12288,128,0.037823110818862915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,128,0.04822044571240743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,64,0.03552444444762336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,64,0.04985510971811083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,12288,32,0.03664177656173706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,12288,32,0.05031377739376492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,65536,8.183609856499567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,65536,4.536144680447049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,1536,0.02251111136542426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,65536,4.722532484266493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,16384,2.029595481024848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,16384,1.0085075166490343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,16384,1.0825564066569011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,12288,0.7737181981404623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,12288,1.4370453092787001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,12288,0.8026302125718859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,10240,0.6480275789896647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,10240,1.181282679239909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,10240,0.6423164473639594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,8192,0.4973920186360677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,8192,0.9067938062879773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,8192,0.5347733497619629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,7168,0.8378053241305881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,1024,0.016887111796273124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,7168,0.4579457706875271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,7168,0.4425306585099962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,6144,0.7026569048563639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,6144,0.37952976756625706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,6144,0.3815404574076335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,5120,0.31357155905829537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,768,0.014078223043017916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,5120,0.5965138011508518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,5120,0.3097733391655816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,4096,0.2583564387427436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,4096,0.45402664608425564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,4096,0.24274399545457628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,3584,0.23121688101026747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,3584,0.22082222832573783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,3584,0.40768888261583114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,3072,0.341048002243042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,3072,0.1930826637479994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,3072,0.1904391050338745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,2560,0.29244976573520237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,2560,0.15571022033691406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,2560,0.16750044292873809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,2048,0.1352248854107327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,2048,0.12850399812062582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,2048,0.23934843805101183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,1536,0.1796355512407091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,1536,0.10565511385599773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,1536,0.10143022404776679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,1024,0.12724622090657553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,1024,0.07677066988415189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,1024,0.07715021901660495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,768,0.0975866648885939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,768,0.06629955768585205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,512,0.07256444295247395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,768,0.06173066960440742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,512,0.04667911264631483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,512,0.05568355321884155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,256,0.03401333424780104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,256,0.04660711023542616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,256,0.045271111859215625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,10240,128,0.03245866629812453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,128,0.03393422232733832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,128,0.04138044516245524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,64,0.030255999830034044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,64,0.04283110962973701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,10240,32,0.031326221095191106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,10240,32,0.043281777037514575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,65536,3.676849789089627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,65536,6.559651692708333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,65536,3.7074606153700085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,16384,0.8072942097981771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,16384,1.5909644232855902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,16384,0.8920835918850369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,12288,1.0861262215508354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,12288,0.6568889088100857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,10240,0.5265768898857964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,12288,0.611645327674018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,10240,0.9041688707139758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,10240,0.5014950964185927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,8192,0.7219715648227267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,8192,0.4105662239922418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,8192,0.43057865566677517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,7168,0.36568178070916074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,7168,0.6157520082261827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,7168,0.3584444522857666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,6144,0.3091324435340034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,6144,0.5412409040662978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,6144,0.31254222657945424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,5120,0.25068532096015084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,5120,0.44131822056240505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,4096,0.3693555461035834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,5120,0.25572178098890513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,4096,0.2109982172648112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,4096,0.19893511136372885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,3584,0.18413421842787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,3584,0.32456088066101074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,3584,0.1832808918423123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,3072,0.16083999474843344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,3072,0.14997243881225586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,3072,0.2742577658759223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,2560,0.13873867193857828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,2560,0.22613689634535047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,2560,0.12786666552225748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,2048,0.11185422208574082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,2048,0.18548177348242867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,2048,0.10637422402699788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,1536,0.14256800545586482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,1536,0.08071466949250963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,1536,0.08728000190522935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,1024,0.06307111183802287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,1024,0.10122488604651557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,1024,0.06300710969501071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,768,0.07778400182723999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,768,0.05365688933266533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,768,0.050853331883748375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,512,0.03933511177698771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,512,0.05766133467356364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,512,0.04555911156866285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,256,0.028676443629794653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,256,0.035981333918041654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,256,0.037448889679378934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,128,0.0277528895272149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,128,0.03433510992262099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,64,0.024496888120969135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,64,0.03549422158135308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,8192,128,0.026986666851573523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,8192,32,0.026131555438041687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,8192,32,0.03580177823702494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,65536,3.2383751339382596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,65536,3.3135867648654513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,16384,1.3489706251356337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,65536,5.720096164279514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,16384,0.7181617948744031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,16384,0.7538666725158691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,12288,0.5478409131368002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,12288,0.9400178061591254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,12288,0.5803795390658909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,10240,0.784345785776774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,10240,0.4521946907043457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,10240,0.48105335235595703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,8192,0.36444889174567324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,8192,0.3714257876078288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,8192,0.6366764704386393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,7168,0.5475102000766331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,7168,0.31486577457851833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,7168,0.3256862163543701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,6144,0.4785617722405328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,6144,0.273780451880561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,6144,0.27060890197753906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,5120,0.22761334313286674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,5120,0.39461776945326066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,5120,0.22719287872314453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,4096,0.31185245513916016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,4096,0.1876222160127428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,4096,0.18339999516805014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,3584,0.2843208842807346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,3584,0.16559110747443304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,3584,0.16127288341522217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,3072,0.14341421922047934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,3072,0.13881511158413357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,3072,0.23726932207743326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,2560,0.19750577873653838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,2560,0.1156160036722819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,512,0.011589333415031433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,2560,0.12186577585008408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,2048,0.10153600242402817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,2048,0.1600817839304606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,2048,0.09298044443130493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,1536,0.12424177593655056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,1536,0.07404889000786676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,1024,0.0867991116311815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,1024,0.055480887492497764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,1536,0.07843377855088976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,768,0.06959288650088839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,768,0.048785779211256236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,1024,0.05725066529379951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,512,0.03583022289805942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,768,0.04544533292452494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,512,0.05101422137684292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,512,0.04158755474620395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,256,0.03250311149491204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,256,0.03455200129085117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,256,0.026534222894244727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,128,0.02490577763981289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,128,0.031488001346588135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,7168,128,0.02404088940885332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,64,0.023288889063729182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,64,0.032214221027162336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,7168,32,0.02377066678471035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,7168,32,0.03228977653715346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,65536,2.923000971476237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,16384,0.6744782129923502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,16384,1.1841724183824327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,65536,4.854576110839844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,65536,2.8027449713812937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,16384,0.6215715408325195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,12288,0.47844621870252824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,12288,0.8481724527147082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,12288,0.4912942250569661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,10240,0.3888568878173828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,10240,0.6667688687642416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,10240,0.40190754996405703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,8192,0.536444452073839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,8192,0.32060088051689994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,8192,0.32512354850769043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,7168,0.2752942244211833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,7168,0.47567023171318906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,7168,0.27867555618286133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,6144,0.4115155537923177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,6144,0.23932178815205893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,6144,0.2384577857123481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,5120,0.1999839941660563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,5120,0.3389635615878635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,5120,0.1953564484914144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,4096,0.16014133559332952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,4096,0.273888005150689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,4096,0.16511021720038518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,3584,0.14417155583699545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,3584,0.24040444691975912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,3584,0.1408053371641371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,3072,0.20718133449554443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,3072,0.12277421686384414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,3072,0.12141333685980903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,2560,0.1070222192340427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,2560,0.1733920044369168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,2560,0.10034222073025173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,2048,0.13798756069607204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,2048,0.08739288647969563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,2048,0.08341866731643677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,1536,0.1080933345688714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,1536,0.0687733358807034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,1536,0.06432177623112996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,1024,0.04974577824274699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,1024,0.07621600230534871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,1024,0.049682666858037315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,768,0.04035377833578322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,768,0.05996711386574639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,768,0.04351822204060025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,512,0.04447466797298855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,512,0.03199377655982971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,512,0.037019554111692644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,256,0.023363555471102398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,256,0.02895911203490363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,256,0.03101244568824768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,6144,128,0.02106933295726776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,128,0.021609778205553692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,128,0.027777777777777776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,64,0.020296888218985665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,64,0.02872266703181797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,6144,32,0.02121066715982225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,6144,32,0.028368890285491943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,65536,2.390397389729818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,65536,4.156656901041667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,12288,0.3942471080356174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,16384,0.5674506823221842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,16384,0.914189338684082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,65536,2.47816891140408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,16384,0.5267537964714898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,12288,0.6837760077582465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,12288,0.4117120107014974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,10240,0.3363786538441976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,10240,0.5624160236782497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,10240,0.3529004520840115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,8192,0.26860088772243923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,8192,0.45778311623467344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,8192,0.2796773380703396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,7168,0.23826755417717826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,7168,0.3885217772589789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,7168,0.2420639991760254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,6144,0.20348178015814888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,6144,0.34555376900566953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,6144,0.20406666066911486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,5120,0.28009867668151855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,5120,0.16844444804721406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,5120,0.17253688971201578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,4096,0.13998044861687556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,4096,0.23386043972439238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,4096,0.13674399587843153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,3584,0.12112622790866429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,3584,0.20255555046929252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,3584,0.1238293382856581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,3072,0.10795288615756565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,3072,0.17787555853525797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,256,0.00814488861295912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,3072,0.10537422365612453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,2560,0.1476960049735175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,2560,0.09189866648779975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,2560,0.08668977684444851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,2048,0.12139733632405598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,2048,0.07165422042210896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,2048,0.07651111152436998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,1536,0.05936444467968411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,1536,0.091867552863227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,1536,0.055663108825683594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,1024,0.04382755690150791
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,1024,0.06492000155978732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,1024,0.042381332980261914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,768,0.03539199961556329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,768,0.0513626668188307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,768,0.03801777627733018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,512,0.03923111160596212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,512,0.03209422363175286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,512,0.028393778536054824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,256,0.021299555897712708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,256,0.02644888891114129
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,256,0.027327110370000202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,128,0.019884443945354886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,128,0.024528889192475214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,64,0.016125332978036668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,5120,128,0.018467555443445843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,64,0.0251555558707979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,5120,32,0.01703644461101956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,5120,32,0.024995555480321247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,65536,3.242960823906792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,65536,1.924658669365777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,65536,2.029362572564019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,16384,0.7331537670559354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,16384,0.4700311024983724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,16384,0.4366773234473334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,12288,0.3240266640981038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,12288,0.5298426416185167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,12288,0.3344053427378337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,10240,0.43852000766330296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,10240,0.27681422233581543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,10240,0.2905200057559543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,8192,0.22268356217278373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,8192,0.34976532724168563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,8192,0.22890133327907988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,7168,0.3097902139027913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,7168,0.19877333111233184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,7168,0.19491110907660592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,6144,0.26862311363220215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,6144,0.16498488850063747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,6144,0.17101511690351698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8192,128,128,0.00618577789929178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,5120,0.2240053282843696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,5120,0.1366355551613702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,5120,0.14190399646759033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,4096,0.1822213331858317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,4096,0.11588977442847358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,4096,0.10891999800999959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,3584,0.16181333859761557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,3584,0.10377333561579387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,3584,0.09664889176686604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,3072,0.13813155227237278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,3072,0.08343199888865153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,3072,0.08936355511347453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,2560,0.07722666528489855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,2560,0.11489333046807183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,2560,0.07177778085072835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,2048,0.06361599763234456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,2048,0.09395111269421047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,2048,0.05901422103246053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,1536,0.049901333120134145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,1536,0.07341511381997003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,1536,0.04692444536421034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,1024,0.052588442961374916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,1024,0.03710133168432448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,1024,0.03552533189455668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,768,0.04188088907135857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,768,0.030590222941504583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,768,0.03166044420666165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,512,0.03150044547186957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,512,0.02442666722668542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,512,0.026153778036435444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,256,0.018460444278187223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,256,0.02159822152720557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,256,0.02181422213713328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,4096,128,0.016335111525323655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,128,0.016743999388482835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,128,0.019359111785888672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,64,0.013638222383128272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,64,0.019645333290100098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,4096,32,0.014190221826235453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,4096,32,0.019644444187482197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,65536,1.972898695203993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,65536,2.9577492607964406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,65536,1.7897101508246527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,16384,0.4383813275231256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,16384,0.680973317888048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,16384,0.41411201159159344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,12288,0.5077973471747504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,12288,0.3046391010284424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,10240,0.41464000278049046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,12288,0.3285111056433784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,10240,0.25031288464864093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,10240,0.27788088056776256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,8192,0.22556532753838432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,8192,0.33436176511976456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,8192,0.2074648936589559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,7168,0.29069333606296116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,7168,0.19784088929494223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,7168,0.17511910862392852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,6144,0.2494115564558241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,6144,0.17033510737948945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,6144,0.15029777420891657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,5120,0.12381954987843831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,5120,0.21030488279130724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,4096,0.11796532736884223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,5120,0.14572089248233372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,4096,0.16728711128234863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,4096,0.09962489207585652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,3584,0.15017599529690212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,3584,0.08852089113659328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,3584,0.10535555415683323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,3072,0.09054666757583618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,3072,0.13088622358110216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,3072,0.07698222001393636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,2560,0.1084959970580207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,2560,0.07819288969039917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,2560,0.06568177541097005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,2048,0.06394577688641019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,2048,0.08824977609846328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,2048,0.05428622166315714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,1536,0.050829334391487964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,1536,0.06953244739108615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,1536,0.04358311163054573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,1024,0.04934666554133097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,1024,0.033081776565975614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,1024,0.04000799854596456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,768,0.03940355446603563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,768,0.03168088859981961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,768,0.029475556479560003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,512,0.025084444218211707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,512,0.02568977740075853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,512,0.023945776952637568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,256,0.019115555617544387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,256,0.0170133329100079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,256,0.020119110743204754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,128,0.014615111880832247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3584,128,0.01695644524362352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,128,0.01789688898457421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,64,0.0129457778400845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,64,0.017729777428838942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3584,32,0.013870221873124441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3584,32,0.018038221531444125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,65536,1.488375981648763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,65536,2.494480980767144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,65536,1.5530266231960719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,16384,0.34508623017205137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,16384,0.5390968852572972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,16384,0.35407378938462997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,12288,0.39378754297892254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,12288,0.26627998881869847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,12288,0.2603493266635471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,10240,0.22102133433024088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,10240,0.22204800446828207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,10240,0.3283653259277344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,8192,0.2697368992699517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,8192,0.1794337828954061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,8192,0.17805422676934135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,7168,0.23571377330356172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,7168,0.1556560066011217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,7168,0.15617777241600886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,6144,0.13533511426713732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,6144,0.20000622007581922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,6144,0.13239199585384792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,5120,0.16834755738576254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,5120,0.10975288682513767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,5120,0.1147315502166748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,4096,0.09292444255616929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,4096,0.13944178157382542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,4096,0.08891910976833767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,3584,0.08333866463767158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,3584,0.12212088372972275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,3584,0.0781244436899821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,3072,0.10639822483062744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,3072,0.06833777825037639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,3072,0.07230133480495878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,2560,0.06284711096021864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,2560,0.08789422114690144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,2560,0.0587351123491923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,2048,0.051817778084013194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,2048,0.07125600179036458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,2048,0.048700445228152804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,1536,0.04096444447835287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,1536,0.05598488781187269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,1536,0.03873599900139703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,1024,0.04060533311631944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,1024,0.03070666723781162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,1024,0.029488000604841445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,768,0.0325617790222168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,768,0.02652799917591943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,512,0.024919110867712233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,512,0.021544888615608215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,768,0.025174222058720056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,512,0.020806221498383414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,256,0.01756888959142897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,256,0.01570755574438307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,256,0.01792711185084449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,3072,128,0.01351199961370892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,128,0.013367999758985309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,128,0.015644444359673392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,64,0.011181332998805575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,64,0.015850666496488783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,3072,32,0.011633777783976661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,3072,32,0.016206221448050607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,65536,1.9125696818033855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,16384,0.3490462303161621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,65536,1.3277111053466797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,65536,1.4606986575656469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,16384,0.4308942159016927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,16384,0.30712087949117023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,12288,0.32791111204359263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,12288,0.23096444871690539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,10240,0.2786906560262044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,12288,0.26473599010043675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,10240,0.22019467088911268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,10240,0.19593511687384713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,8192,0.22113422552744547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,8192,0.17826933330959746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,8192,0.15412978331247965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,7168,0.15894222259521484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,7168,0.19637155532836914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,7168,0.13463466697269016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,6144,0.17030755678812662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,6144,0.11634222666422527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,6144,0.13615288999345568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,5120,0.11507999897003174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,5120,0.14002044995625815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,5120,0.09586311048931545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,4096,0.11368710464901394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,4096,0.09460533327526516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,4096,0.07783555322223239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,3584,0.08391555812623765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,3584,0.09866488642162746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,3584,0.06891110870573255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,3072,0.08673599693510269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,3072,0.07382222016652425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,3072,0.06018222040600247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,2560,0.07419733206431071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,2560,0.06345155504014757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,2560,0.05180089010132683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,2048,0.05262666609552172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,2048,0.05885510974460178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,2048,0.04309511184692383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,1536,0.04239733351601494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,1536,0.04606577754020691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,1536,0.03412533468670315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,1024,0.03324177861213684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,1024,0.03279822071393331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,1024,0.026375111606385972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,768,0.026312000221676294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,768,0.023939554889996845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,768,0.027469333675172593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,512,0.02215911116864946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,512,0.01995377739270528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,512,0.01938577824168735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,256,0.016180444094869826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,256,0.013618666264745923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,256,0.015967999895413715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,128,0.010216000179449717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2560,128,0.013999111122555204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,128,0.014296889305114746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,64,0.008651555412345463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,64,0.01424711114830441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2560,32,0.009045333498054082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2560,32,0.014459555347760519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,65536,1.1143947177463107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,65536,1.501023080613878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,65536,1.0346683926052518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,16384,0.25327377849155003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,16384,0.35243111186557347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,16384,0.25968978140089244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,12288,0.19194044007195366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,12288,0.2564728789859348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,12288,0.1948853333791097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,10240,0.21967555416954887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,10240,0.16338222556644016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,10240,0.15957244237263998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,8192,0.17752533488803438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,8192,0.13125688499874538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,8192,0.131315549214681
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,7168,0.11580177148183186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,7168,0.15401511722140843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,7168,0.11519466506110297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,6144,0.10003466738594903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,6144,0.13127644856770834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,6144,0.10043733649783665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,5120,0.1116524404949612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,5120,0.08535377846823798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,5120,0.0820008913675944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,4096,0.0905288921462165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,4096,0.07048977745903863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,4096,0.06633955240249634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,3584,0.06175644530190361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,3584,0.08052533202701144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,3584,0.05904089079962837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,3072,0.054551111327277295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,3072,0.0683200028207567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,3072,0.05115466647677951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,2560,0.04748177859518263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,2560,0.058279110325707324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,2560,0.04374577932887607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,2048,0.03941955500178867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,2048,0.04739022254943848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,2048,0.036441776487562395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,1536,0.03675555520587497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,1536,0.031167109807332356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,1536,0.029370665550231934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,1024,0.02606666584809621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,1024,0.024148444334665935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,1024,0.022673777408070032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,768,0.021700445148679946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,768,0.02115999990039402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,512,0.016389333539538913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,768,0.020232889387342665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,512,0.016718222035302054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,512,0.01696622206105126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,256,0.012665777570671506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,256,0.011286222272449069
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,256,0.01372977760103014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,128,0.008444444172912175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,128,0.01203555530971951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,64,0.007270221908887227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,2048,128,0.011081777513027191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,64,0.012595555848545499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,2048,32,0.007640000018808577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,2048,32,0.0125591109196345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,65536,1.1043671502007377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,65536,0.8525848918490939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,65536,0.9025768703884549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,16384,0.20946844418843588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,16384,0.27101778984069824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,16384,0.22491643163892958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,12288,0.20626666810777453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,12288,0.15685510635375977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,12288,0.1615413294898139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,10240,0.1324711110856798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,10240,0.170869337187873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,10240,0.1358684433831109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,8192,0.10767200258043076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,8192,0.13565599918365479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,8192,0.11411021815405951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,7168,0.12006399366590713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,7168,0.09497777620951335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,7168,0.09816444582409328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,6144,0.08273066414727105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,6144,0.10562400023142497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,6144,0.0839537779490153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,5120,0.08828888999091254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,5120,0.0697279969851176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,5120,0.06965155733956231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,4096,0.06946044498019747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,4096,0.057909336354997426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,4096,0.056735111607445605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,3584,0.062241779433356396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,3584,0.05032889048258463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,3584,0.051875554853015475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,3072,0.045951998896069
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,3072,0.0528675549560123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,3072,0.04351822204060025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,2560,0.039352890517976545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,2560,0.04545066754023234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,2560,0.0370435549153222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,2048,0.03772088885307312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,2048,0.03087466624048021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,1536,0.028931554820802476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,2048,0.03306577934159173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,1536,0.026919111609458923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,1536,0.025222222010294598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,1024,0.02093688812520769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,1024,0.02090666691462199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,1024,0.019857777489556205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,768,0.017104888955752056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,768,0.018636445204416912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,768,0.017941332525677152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,512,0.015152888165579902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,512,0.013362666798962487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,512,0.014714666538768344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,256,0.011701333026091257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,256,0.008990222381220924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,256,0.011438222395049201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,128,0.007240888973077138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,128,0.010469333165221743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1536,128,0.009955555200576782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,64,0.006141333530346553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,64,0.010868444210953183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1536,32,0.0064080001579390625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1536,32,0.011015111373530494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,65536,0.7170569101969401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,65536,0.7200622028774686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,65536,0.6502408981323242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,16384,0.17502755588955352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,16384,0.18036177423265246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,16384,0.16577422618865967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,12288,0.12510311603546143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,12288,0.13416443930731878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,12288,0.13097688886854383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,10240,0.10491022136476304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,10240,0.10858400000466241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,10240,0.11117511325412327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,8192,0.09014399846394856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,8192,0.09336711300743951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,8192,0.08626221948199803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,7168,0.07818222045898438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,7168,0.07656000057856242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,7168,0.08307200007968478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,6144,0.06680177979999118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,6144,0.06795289119084676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,6144,0.0698524448606703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,5120,0.05693688657548693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,5120,0.0573671129014757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,5120,0.0573822193675571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,4096,0.047213332520590894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,4096,0.045838223563300244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,4096,0.04638400011592441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,3584,0.041527112325032554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,3584,0.04169866773817274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,3072,0.03587466809484694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,3584,0.04212799999448988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,3072,0.03730488816897074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,3072,0.03548266821437412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,2560,0.030469334787792627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,2560,0.03220444586541917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,2560,0.03068888849682278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,2048,0.025384889708624944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,2048,0.02574933403068119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,1536,0.019966221517986722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,2048,0.027262222435739305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,1536,0.02254933284388648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,1536,0.021391999390390184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,1024,0.017712000343534682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,1024,0.015133332875039844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,1024,0.01667644414636824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,768,0.015152888165579902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,768,0.0125608891248703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,768,0.016185777054892648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,512,0.013175110850069257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,512,0.009711999860074785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,512,0.012629333469602795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,256,0.010495999621020423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,256,0.0070471109615431885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,256,0.009667555491129557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,1024,128,0.0090515555606948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,128,0.0057911111248864075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,128,0.008955555657545725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,64,0.005140444470776452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,64,0.009224889179070791
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,1024,32,0.005335999859703912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,1024,32,0.009381333159075843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,65536,0.5533279842800564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,65536,0.557046201494005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,65536,0.6825813187493218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,16384,0.13439022170172796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,16384,0.1419644488228692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,16384,0.1766888830396864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,12288,0.10819733142852783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,12288,0.10316266616185506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,12288,0.12845688396030003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,10240,0.09088177813424005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,10240,0.08670933379067315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,10240,0.10812533564037746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,8192,0.07443733347786798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,8192,0.06919111145867242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,8192,0.09069777859581842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,7168,0.06635289059744941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,7168,0.061794665124681264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,7168,0.07903289132648043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,6144,0.05764089028040568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,6144,0.052264889081319175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,6144,0.06859999895095825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,5120,0.04911999901135763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,5120,0.04528533418973287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,5120,0.05577511257595486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,4096,0.03700800074471368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,4096,0.04530488782458835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,3584,0.036715557177861534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,4096,0.040735999743143715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,3584,0.032418668270111084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,3584,0.04018311036957635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,3072,0.03237333231502109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,3072,0.028163555595609877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,3072,0.03443377878930833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,2560,0.02803200152185228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,2560,0.024041778511471216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,2560,0.029931555191675823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,2048,0.02397955622937944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,2048,0.019852444529533386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,2048,0.025504888759719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,1536,0.019732443822754752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,1536,0.016064888901180692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,1536,0.02122755514250861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,1024,0.01221422188811832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,1024,0.016492444607946608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,768,0.010489777558379702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,768,0.015843555331230164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,1024,0.015855110353893705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,768,0.013597333596812354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,512,0.011980444192886353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,512,0.008023999631404877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,512,0.012383999923865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,256,0.005959110955397288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,256,0.009880888793203566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,256,0.009600888523790572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,768,128,0.008239111138714684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,128,0.005074666606055366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,128,0.008950221869680617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,64,0.004531555705600315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,64,0.009082666701740688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,768,32,0.004690666579537922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,768,32,0.009213333328564962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,65536,0.481736871931288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,65536,0.3627955648634169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,65536,0.529279973771837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,16384,0.1236835585700141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,16384,0.09656711419423421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,16384,0.1322257783677843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,12288,0.0723902251985338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,12288,0.09376266929838394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,12288,0.09968711270226373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,10240,0.07867377996444702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,10240,0.06321333514319526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,10240,0.08559822373920017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,8192,0.049639112419552274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,8192,0.07076444228490193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,8192,0.06463288598590426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,7168,0.04344444473584493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,7168,0.05746489100986057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,7168,0.062169777022467725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,6144,0.036892443895339966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,6144,0.05006666647063362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,6144,0.0540888872411516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,5120,0.03164533442921109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,5120,0.04447911183039347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,4096,0.03554577628771464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,5120,0.043078223864237465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,4096,0.025472889343897503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,4096,0.03596355517705282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,3584,0.022966222630606756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,3584,0.0314408904976315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,3584,0.0318488876024882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,3072,0.028079112370808918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,3072,0.020273778173658583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,3072,0.027041777968406677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,2560,0.017066667477289837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,2560,0.024680889315075342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,2560,0.023461333579487268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,2048,0.014584889014561972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,2048,0.02049777739577823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,1536,0.011955555942323474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,1536,0.017601778109868366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,2048,0.020947555700937908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,1536,0.01752800080511305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,1024,0.009383111364311641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,1024,0.014163555370436775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,1024,0.013756444056828817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,768,0.012182222472296821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,768,0.00816177742348777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,768,0.013880000346236758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,512,0.00656355544924736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,512,0.010649777948856354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,512,0.010025777750545079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,256,0.005015999906592899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,256,0.007835555407736037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,256,0.008208889100286696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,128,0.004426666845877965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,512,128,0.006552888701359431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,128,0.007582222421964009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,64,0.0039822223285834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,64,0.007713778151406183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,512,32,0.004243555582231945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,512,32,0.007827555967701806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,65536,0.4169928762647841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,65536,0.20138577620188394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,65536,0.4909537633260091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,16384,0.05678488810857137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,16384,0.12354311678144668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,16384,0.10880177550845677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,12288,0.08101155360539754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,12288,0.04393600092993843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,12288,0.09553511275185479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,10240,0.03722755445374383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,10240,0.06944177548090617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,10240,0.08027555545171101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,8192,0.03490488727887472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,8192,0.05746755335066053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,8192,0.0671164459652371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,7168,0.05076533224847582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,7168,0.028044445647133723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,7168,0.05960622098710802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,6144,0.04432622260517544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,6144,0.02382755610677931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,6144,0.0521031121412913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,5120,0.03813955518934462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,5120,0.020830222302012973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,5120,0.042578667402267456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,4096,0.018990221950742934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,4096,0.03131822082731459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,4096,0.03361777795685662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,3584,0.028776887390348647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,3584,0.016548444827397663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,3584,0.02954933378431532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,3072,0.025032889511850145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,3072,0.013793778088357715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,3072,0.025989333788553875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,2560,0.012553777959611682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,2560,0.02292977770169576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,2560,0.02195466723706987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,2048,0.018552889426549275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,2048,0.01019200020366245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,2048,0.02016177773475647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,1536,0.015619556109110514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,1536,0.008960000342792934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,1536,0.0166640000210868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,1024,0.012507555385430654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,1024,0.007174222005738153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,1024,0.013373333546850415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,768,0.010671111444632212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,768,0.006011555592219035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,768,0.013421333498424955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,512,0.008775999976529015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,512,0.004860444615284602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,512,0.009667555491129557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,256,0.0041786668201287585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,256,0.006521777974234686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,256,0.008025777836640676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,128,0.003673777812057071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,128,0.007368889119890001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,256,128,0.005618666609128316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,64,0.0033973333322339584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,64,0.00740444411834081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,256,32,0.0035626664757728577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,256,32,0.007479999628331926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,65536,0.16345243983798557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,65536,0.4867786831325955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,65536,0.4027991029951308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,16384,0.10339199834399754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,16384,0.042753779225879245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,16384,0.12277155452304417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,12288,0.07771466837988959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,12288,0.03404622276624044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,12288,0.0939084423912896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,10240,0.030045333835813735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,10240,0.08023733562893338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,10240,0.06570755773120456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,8192,0.026308443811204698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,8192,0.06658933560053508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,8192,0.05398133397102356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,7168,0.01945955554644267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,7168,0.05927466683917575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,7168,0.04783199893103706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,6144,0.017104888955752056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,6144,0.051775111092461475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,6144,0.041238221857282854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,5120,0.03552177879545424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,5120,0.014932443698247274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,5120,0.041871110598246254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,4096,0.029254221253924902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,4096,0.013272888958454132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,4096,0.03287377622392442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,3584,0.0114720000161065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,3584,0.02889510989189148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,3584,0.026695999834272597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,3072,0.010220444036854638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,3072,0.02345866627163357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,3072,0.025293333662880793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,2560,0.009170666337013245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,2560,0.020594666401545208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,2560,0.022861333356963262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,2048,0.01758044461409251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,2048,0.007992000215583378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,2048,0.01995466649532318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,1536,0.0069751110341813825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,1536,0.014496889379289416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,1536,0.016552888684802584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,1024,0.011680000358157687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,1024,0.005872888697518243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,1024,0.013269333375824822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,768,0.0047848886913723415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,768,0.013241777817408243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,512,0.004054222255945206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,512,0.008780444661776224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,768,0.009757333331637913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,512,0.007740444607204861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,256,0.006024000131421619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,256,0.003588444242874781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,256,0.007800000409285228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,128,0.0033493333806594214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,128,0.007119999991522894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4096,128,128,0.0053173332578606075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,64,0.00309599998096625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,64,0.007227555745177799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,128,32,0.003469333259595765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,128,32,0.007252444823582967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,65536,0.14078222380744085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,16384,0.03960444529851278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,65536,0.4858071009318034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,16384,0.12247467041015625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,12288,0.03267466690805223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,12288,0.09354399972491795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,10240,0.02861600120862325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,10240,0.08014400137795342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,8192,0.024837333295080397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,8192,0.06707377566231622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,7168,0.022674666510687932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,7168,0.05906755394405789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,6144,0.019766221443812054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,6144,0.05158399873309665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,5120,0.014222222897741528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,5120,0.041617777612474226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,4096,0.012524444195959302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,4096,0.032812443044450544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,3584,0.010858666565683154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,3584,0.028809779220157202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,3072,0.009745777481132084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,3072,0.025278222229745653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,2560,0.008813333180215623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,2560,0.02271466619438595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,2048,0.007831110722488826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,2048,0.019753777318530612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,1536,0.006904888898134232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,1536,0.016319110989570618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,1024,0.0058106668293476105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,1024,0.013191111385822296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,768,0.00444177786509196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,768,0.013140444126394061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,512,0.0038240001433425476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,512,0.008831111093362173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,256,0.0034168888297345904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,256,0.007801777786678738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,128,0.0031226666437255014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,128,0.007143999967310164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,64,0.0028693332440323303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,64,0.007155555817815993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,64,32,0.003119999956753519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,64,32,0.007199111084143321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,65536,0.1358755562040541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,16384,0.037951111793518066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,16384,0.12246044476826985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,65536,0.4839635425143772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,12288,0.031818668047587075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,12288,0.09350399838553534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,10240,0.02767555581198798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,10240,0.08027288648817274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,8192,0.02312799957063463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,8192,0.0665902230474684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,7168,0.021176000436147053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,7168,0.05899733304977417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,6144,0.01884888940387302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,6144,0.05155466662512886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,5120,0.016164445214801364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,5120,0.04165600074662103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,4096,0.014117333624098035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,4096,0.03269511130121019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,3584,0.012832889126406776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,3584,0.028590222199757893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,3072,0.011800888511869641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,3072,0.02532711128393809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,2560,0.008665777742862701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,2560,0.022664889693260193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,2048,0.007781333393520779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,2048,0.019849777221679688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,1536,0.006673777682913675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,1536,0.016386666231685214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,1024,0.005752888818581899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,1024,0.013234666652149625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,768,0.0041288890772395665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,768,0.012903110848532783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,512,0.0036799998746977914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,512,0.009043555292818282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,256,0.0032595555401510666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,256,0.007794666621420119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,128,0.002962666667169995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,128,0.007150222029950883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,64,0.0027946666296985415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,64,0.007197333706749811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4096,32,32,0.0030782222747802734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4096,32,32,0.007204444044166141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,16384,3.472487131754557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,16384,6.569478352864583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,16384,3.4551448822021484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,12288,2.573613272772895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,12288,4.909272935655382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,12288,2.5774623023139105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,10240,2.1795946756998696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,10240,4.0856781005859375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,10240,2.112303204006619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,8192,1.7314568625556097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,8192,3.3429431915283203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,8192,1.6245706346299913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,7168,1.4739031261867945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,7168,2.900083541870117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,7168,1.404319127400716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,6144,1.2290177875094943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,6144,2.5050392150878906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,6144,1.2750977410210502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,5120,0.9892507129245334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,5120,2.0487325456407333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,5120,0.9996666378445096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,4096,0.7579626507229276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,4096,1.6573973761664496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,4096,0.7826808823479546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,3584,0.7168275515238444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,3584,1.3984248903062608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,3584,0.662043571472168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,3072,1.2053475909762912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,2560,0.4968871010674371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,3072,0.5740506913926866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,2560,0.4739715788099501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,2560,0.9954142040676541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,3072,0.5923546685112847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,2048,0.7627155515882703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,2048,0.40931910938686794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,2048,0.37325867017110187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,1536,0.5681644545661079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,1536,0.3158400058746338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,1536,0.29692355791727704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,1024,0.3825333383348253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,1024,0.2261422210269504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,1024,0.2284888956281874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,768,0.18040800094604492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,768,0.2994604375627306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,768,0.18827911218007407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,512,0.21867999765608045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,512,0.1563591162363688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,512,0.13402399751875135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,256,0.1371119949552748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,256,0.13103200329674616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,256,0.10249244504504734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,65536,128,0.10067111253738403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,128,0.1066266695658366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,128,0.12232089042663574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,64,0.09516888856887817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,64,0.12701688872443304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,65536,32,0.09415733151965672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,65536,32,0.12645778391096327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,65536,6.474493238661025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,65536,3.507271024915907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,65536,3.568055046929253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,16384,0.826656871371799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,16384,0.8226488961113824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,16384,1.5238996081882055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,12288,0.5887040032280816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,12288,1.093909369574653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,12288,0.5699466599358453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,10240,0.5006951226128472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,10240,0.9420666164822049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,10240,0.4899431334601508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,8192,0.3806302282545302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,8192,0.7373751004536947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,8192,0.39075289832221133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,7168,0.3451288806067572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,7168,0.618975109524197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,6144,0.2968382305569119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,7168,0.33419020970662433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,6144,0.5467271274990505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,6144,0.28669510947333443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,5120,0.4515226682027181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,5120,0.24517067273457846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,5120,0.237373325559828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,4096,0.19771022266811797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,4096,0.3654630978902181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,4096,0.1950248877207438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,3584,0.1683511071734958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,3584,0.32385065820482045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,3072,0.27674134572347003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,3584,0.17759377426571318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,3072,0.15259110927581787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,3072,0.14443733957078722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,2560,0.12349511517418756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,2560,0.23555644353230795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,2048,0.10957866907119751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,2048,0.18720711602105033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,2048,0.10177244742711385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,2560,0.1322373284233941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,1536,0.08454489045672947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,1536,0.1413875553343031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,1536,0.079857779873742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,1024,0.06228799952401055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,1024,0.10119911034901936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,1024,0.0617920027838813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,768,0.07913511329227023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,768,0.04958933260705736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,768,0.052777777115503945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,512,0.03866577810711331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,512,0.057504885726504855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,512,0.046923554605907865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,256,0.029791110091739233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,256,0.03625422385003831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,256,0.040645334455702044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,16384,128,0.0270453327231937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,128,0.028635554843478735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,128,0.03757955630620321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,64,0.025356445047590468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,64,0.03895022140608893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,16384,32,0.025978666212823655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,16384,32,0.039020445611741804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,65536,2.6720932854546438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,65536,4.896311865912543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,65536,2.619540532430013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,16384,0.5986106660630968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,16384,1.078627586364746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,16384,0.5803066889444987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,12288,0.4476568963792589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,12288,0.7989928987291125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,12288,0.4478178024291992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,10240,0.37255022260877824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,10240,0.37112445301479763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,10240,0.6568249066670736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,8192,0.5406568845113119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,8192,0.29003998968336314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,7168,0.26319556766086155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,8192,0.2972968949211968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,7168,0.4757466846042209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,7168,0.252222220102946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,6144,0.3975653383466933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,6144,0.22900178697374132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,6144,0.21368355221218535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,5120,0.18121777640448677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,5120,0.33975733651055234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,5120,0.18936533398098418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,4096,0.15501956144968668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,4096,0.274653328789605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,4096,0.14397866196102566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,3584,0.13822933038075766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,3584,0.23737777603997123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,3584,0.12950578000810412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,3072,0.11823022365570068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,3072,0.20714933342403838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,3072,0.11420533392164443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,2560,0.171287112765842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,2560,0.10164088673061794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,2560,0.09795733292897542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,2048,0.08263644244935778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,2048,0.14017066690656874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,2048,0.0788924429151747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,1536,0.1095048917664422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,1536,0.06176800198025174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,1024,0.07738666401969062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,1024,0.04782755507363213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,1536,0.06578399737675984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,1024,0.04845244354671902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,768,0.03958933220969306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,768,0.06037866406970554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,768,0.042423999971813627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,512,0.044863998889923096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,512,0.030976888206270006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,512,0.0379671123292711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,256,0.02405244443151686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,256,0.0292577776643965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,256,0.03288177649180094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,12288,128,0.021154666940371197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,128,0.02215733296341366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,128,0.03066666589842902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,64,0.020488000578350492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,64,0.03160088923242357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,12288,32,0.021928000781271193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,12288,32,0.03128000100453695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,65536,4.139193640814887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,65536,2.2281680636935763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,16384,0.48925420973036027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,16384,0.9580106735229492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,65536,2.2385732862684464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,16384,0.5149484210544163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,12288,0.3737173345353868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,12288,0.7053333388434516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,12288,0.3723768923017714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,10240,0.3013288974761963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,10240,0.5536222457885742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,10240,0.31153866979810924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,8192,0.243797328737047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,8192,0.46373690499199766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,7168,0.20788089434305826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,7168,0.3971840010748969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,6144,0.33942490153842503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,6144,0.18340088261498344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,5120,0.28114133410983616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,5120,0.15439466635386148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,4096,0.2308248943752713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,8192,0.25502310858832467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,7168,0.22249955601162383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,6144,0.18786222404903838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,4096,0.12215199735429551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,5120,0.16273421711391872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,4096,0.13060800234476724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,3584,0.2034906678729587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,3584,0.1113742192586263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,3584,0.11582222249772813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,3072,0.17326666249169242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,3072,0.09436266952090794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,3072,0.10054044591055976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,2560,0.14743288358052573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,2560,0.08729422092437744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,2560,0.08147733079062568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,2048,0.07168710894054837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,2048,0.12089778317345513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,2048,0.06627733177608915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,1536,0.0564862224790785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,1536,0.09293688668145074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,1536,0.05347999930381775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,1024,0.041821334097120494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,1024,0.06524533033370972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,1024,0.04169333312246534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,768,0.034147553973727755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,768,0.051473776499430336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,768,0.036920001109441124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,512,0.027242667145199243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,512,0.039162665605545044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,512,0.033217777808507286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,256,0.020986666282018025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,256,0.026483555634816486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,256,0.028713779317008123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,10240,128,0.01844800015290578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,128,0.020062221421135795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,128,0.026742221580611333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,64,0.016159999701711867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,64,0.02743911080890232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,10240,32,0.01714666684468587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,10240,32,0.027294221851560805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,65536,1.7878470950656469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,65536,3.2457076178656687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,65536,1.7633778254191081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,16384,0.7450702455308703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,16384,0.4065706729888916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,16384,0.40737154748704696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,12288,0.3041324350568983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,12288,0.5540329085456001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,12288,0.29937865999009877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,10240,0.25353066126505536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,10240,0.24684177504645455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,10240,0.4381297694312201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,8192,0.20675378375583223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,8192,0.36419023407830137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,8192,0.1979457802242703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,7168,0.18300267060597739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,7168,0.3146337668100993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,7168,0.16944711738162568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,6144,0.1562026606665717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,6144,0.2660728825463189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,6144,0.14905422263675266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,5120,0.22686844401889375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,5120,0.12668089071909586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,5120,0.1332319974899292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,4096,0.18077600002288818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,4096,0.10044355524910821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,3584,0.09568711121877034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,3584,0.1622755527496338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,4096,0.10699199967914158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,3584,0.0896400014559428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,3072,0.1423084470960829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,3072,0.08323999908235338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,3072,0.07726755407121447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,2560,0.0719662242465549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,2560,0.1160622172885471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,2560,0.06664266851213244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,2048,0.09526222281985813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,2048,0.0589911142985026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,2048,0.05573599868350559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,1536,0.07250577873653836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,1536,0.04735111196835836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,1536,0.044178667995664805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,1024,0.035255998373031616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,1024,0.052611556318071157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,1024,0.034734222624037005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,768,0.042196445994906955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,768,0.029190222422281902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,768,0.0309840010272132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,512,0.023370666636361018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,512,0.03161333335770501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,512,0.027862221002578735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,256,0.018301332990328472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,256,0.021623111433453027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,256,0.024119999673631456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,8192,128,0.01572711103492313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,128,0.016456888781653512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,128,0.022485334012243483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,64,0.013602666556835175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,64,0.022460444105996027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,8192,32,0.014639111028777229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,8192,32,0.022487999664412603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,65536,1.8491342332628038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,65536,2.8076150682237415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,65536,1.5773671468098958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,16384,0.348543988333808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,16384,0.6376346482170953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,12288,0.3093449009789361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,16384,0.4127955436706543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,12288,0.4709911346435547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,12288,0.26943643887837726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,10240,0.2587289015452067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,10240,0.21645511521233451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,10240,0.3836177719963922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,8192,0.3154462178548177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,8192,0.17593510945638022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,8192,0.20741689205169678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,7168,0.18288800451490614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,7168,0.27955378426445854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,7168,0.1528657807244195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,6144,0.1573022206624349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,6144,0.23271377881368002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,6144,0.13303644127315944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,5120,0.13393867015838623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,5120,0.1961688862906562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,5120,0.1124453279707167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,4096,0.1563004387749566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,4096,0.09037777450349595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,4096,0.109516441822052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,3584,0.1397848924001058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,3584,0.09806222385830349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,3584,0.08105066749784681
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,3072,0.0845795538690355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,3072,0.11878132820129395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,3072,0.06971555285983615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,2560,0.07374578052096896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,2560,0.10102933645248413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,2560,0.05992444356282552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,2048,0.06029599905014038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,2048,0.08115110794703166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,2048,0.04980533321698507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,1536,0.047684444321526416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,1536,0.06067022350099352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,1536,0.040044443474875555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,1024,0.037964443365732826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,1024,0.04323644439379374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,1024,0.03180177675353156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,768,0.030339555607901678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,768,0.03380266825358073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,768,0.02831288841035631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,512,0.025124443901909724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,512,0.02445777753988902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,512,0.025243555506070454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,256,0.016184000505341422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,256,0.02195555633968777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,256,0.01903288894229465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,7168,128,0.01676266723208957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,128,0.01350933313369751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,128,0.020064888728989493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,64,0.012587555580668978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,64,0.01938488913906945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,7168,32,0.01350400017367469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,7168,32,0.01886755559179518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,65536,1.3274978001912434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,65536,2.4822213914659286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,65536,1.3340088526407878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,16384,0.30784532758924693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,16384,0.3128524356418186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,16384,0.5332541995578343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,12288,0.4043324523501926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,12288,0.23119465510050455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,10240,0.19693599806891549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,12288,0.2316275437672933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,10240,0.18826399909125435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,10240,0.3348960081736247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,8192,0.16180266274346247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,8192,0.2691440052456326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,8192,0.15339467260572645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,7168,0.23542221387227377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,7168,0.1408240000406901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,7168,0.13282044728597006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,6144,0.12125333150227864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,6144,0.20354578230116102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,6144,0.1176275544696384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,5120,0.17285688718159994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,5120,0.10251200199127197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,5120,0.0983840028444926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,4096,0.0836942195892334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,4096,0.13649155033959284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,4096,0.07839466465844049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,3584,0.12401155630747478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,3584,0.07490311066309611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,3584,0.0709199971622891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,3072,0.06496089034610324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,3072,0.10529333353042603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,3072,0.061512000030941434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,2560,0.08821333116955227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,2560,0.053449776437547475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,2048,0.07311911053127713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,2048,0.04368711180157132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,2560,0.056087109777662486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,2048,0.047147555483712085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,1536,0.037880887587865196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,1536,0.05632088581720988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,1536,0.03584444522857666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,1024,0.04063733418782552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,1024,0.02853333287768894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,1024,0.028768890433841284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,768,0.03286133209864298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,768,0.02551377813021342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,768,0.02368800010946062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,512,0.019519999623298645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,512,0.025007999605602686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,512,0.022348443667093914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,256,0.015077332655588785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,256,0.01756533318095737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,256,0.019174221489164565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,6144,128,0.013277333643701343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,128,0.013510222236315409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,128,0.017163554827372234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,64,0.01128088848458396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,64,0.017222222354676988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,6144,32,0.011307555768224927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,6144,32,0.017463111215167575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,65536,2.0409795973036027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,65536,1.3617404301961262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,65536,1.092808935377333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,16384,0.31424087948269314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,16384,0.2630506621466743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,16384,0.4380151165856256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,12288,0.23575199974907768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,12288,0.3397635618845622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,12288,0.19391200277540419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,10240,0.1983475552664863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,10240,0.27558665805392796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,10240,0.1636435588200887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,8192,0.2211271127065023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,8192,0.1307742198308309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,8192,0.16107022762298584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,7168,0.1428906652662489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,7168,0.1966115501191881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,7168,0.11444532871246338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,6144,0.17056443956163195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,6144,0.09876622094048394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,6144,0.12137244807349311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,5120,0.144176893764072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,5120,0.08516355355580647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,4096,0.08485866917504205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,4096,0.11561066574520534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,5120,0.10308533244662815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,4096,0.06767911381191678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,3584,0.07685244745678373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,3584,0.10001599788665771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,3584,0.06088711155785454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,3072,0.06597689125272962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,3072,0.08758222394519383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,3072,0.053528000911076866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,2560,0.05819377634260389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,2560,0.07442133294211493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,2560,0.04593066705597771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,2048,0.06094666984346178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,2048,0.038239998949898615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,1536,0.03862844573126899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,1536,0.04666577776273092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,2048,0.04784977767202589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,1536,0.03130133284462823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,1024,0.03092799915207757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,1024,0.033040000332726374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,1024,0.025260445144441392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,768,0.02500444485081567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,768,0.026505778233210247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,768,0.022395556171735127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,512,0.02126844392882453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,512,0.01994755533006456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,512,0.019848000672128465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,256,0.015761777758598328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,256,0.013730666703648038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,256,0.01705422169632382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,5120,128,0.013910222384664746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,128,0.010288000106811523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,128,0.014956444501876831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,64,0.008783111141787635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,64,0.015206222732861837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,5120,32,0.008986666798591614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,5120,32,0.015386665860811869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,65536,1.5193039576212566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,65536,0.9555768966674805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,65536,0.9341582192314996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,16384,0.351381328370836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,16384,0.21533689233991835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,16384,0.21669688489702013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,12288,0.26693420939975315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,12288,0.15948443942599827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,12288,0.16611644956800672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,10240,0.14196977350446913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,10240,0.22131910589006212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,10240,0.1343511078092787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,8192,0.11347111066182454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,8192,0.1754364437527127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,8192,0.10781688822640313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,7168,0.10140533579720391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,7168,0.15361066659291586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,7168,0.09586222304238214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,6144,0.08741422494252522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,6144,0.1349466641743978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,6144,0.08233333296246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,5120,0.11413778199089898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,5120,0.07348444726732042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,5120,0.07113689184188843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,4096,0.06138222085105049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,4096,0.0929795569843716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,4096,0.056603557533688016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,3584,0.05452444487147861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,3584,0.08080889119042291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,3584,0.05073422193527222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,3072,0.047242667939927846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,3072,0.06960266828536987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,3072,0.0447377794318729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,2560,0.041703999042510986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,2560,0.058667553795708545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,2560,0.0384115543630388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,2048,0.03496177660094367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,2048,0.04843022094832527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,2048,0.03205777870284186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,1536,0.037069333924187556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,1536,0.026662222213215295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,1024,0.02632444434695774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,1024,0.0217884447839525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,1536,0.028776887390348647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,1024,0.022265776991844177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,768,0.018808888064490426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,768,0.021334222621387903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,768,0.019612444771660697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,512,0.016383111476898193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,512,0.017301332619455125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,512,0.015779554843902588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,256,0.01256711118751102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,256,0.011410666836632622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,256,0.014538667268223233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,4096,128,0.011184888581434885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,128,0.00853066643079122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,128,0.013003555436929068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,64,0.007248889240953658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,64,0.013409777647919126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,4096,32,0.007723555796676212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,4096,32,0.01349066694577535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,16384,0.22267021073235405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,65536,0.8862631056043836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,65536,0.8671360015869141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,65536,1.4622435039944117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,16384,0.21160622437795004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,16384,0.33319200409783256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,12288,0.2516293260786268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,12288,0.15628977616628012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,12288,0.16866489251454672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,10240,0.2064488861295912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,10240,0.14242755042182073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,10240,0.13219822777642146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,8192,0.11611466937594944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,8192,0.16872089438968232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,8192,0.10890844133165146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,7168,0.10158666637208726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,7168,0.1478995614581638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,7168,0.09291999869876438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,6144,0.12629332807328966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,6144,0.08892889155281915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,6144,0.08157777786254883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,5120,0.10730666584438747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,5120,0.06939822435379028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,5120,0.07508266634411283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,4096,0.08563555611504449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,4096,0.0622311102019416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,4096,0.0553288890255822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,3584,0.05634311172697279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,3584,0.07798755831188627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,3584,0.049754665957556836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,3072,0.06613244613011678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,3072,0.04856444398562113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,3072,0.043987555636299976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,2560,0.04336444536844889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,2560,0.05542844533920288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,2560,0.037920888927247785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,2048,0.035875555541780256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,2048,0.04564355479346382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,2048,0.03194133440653483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,1536,0.02902311086654663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,1536,0.035664889547559954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,1536,0.026616000466876562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,1024,0.023767110374238756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,1024,0.02574933403068119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,1024,0.021731555461883545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,768,0.01959111127588484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,768,0.020979555116759405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,768,0.019581332802772522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,512,0.016504888733228047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,512,0.016350222958458793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,512,0.017384888397322763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,256,0.011133333047231039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,256,0.012716444830099741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,256,0.014333332578341166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,128,0.008273777862389883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3584,128,0.011618666350841522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,128,0.013039111263222165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,64,0.007124444676770105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,64,0.01332800007528729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3584,32,0.007646222081449296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3584,32,0.013336000343163809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,65536,0.700984901852078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,65536,0.7233689096238879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,65536,1.2147484885321724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,16384,0.2731715573204888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,16384,0.16795999473995635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,16384,0.17325688732994926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,12288,0.20855732758839926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,12288,0.13433867030673557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,12288,0.12655377388000488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,10240,0.11087377866109212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,10240,0.17293066448635527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,10240,0.10652177863650852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,8192,0.13641688558790419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,8192,0.08537333541446263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,8192,0.08968089024225871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,7168,0.08039999670452543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,7168,0.12196266651153564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,7168,0.0762631098429362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,6144,0.10484177536434597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,6144,0.06670222017500135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,5120,0.08975822395748562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,5120,0.05741600195566813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,6144,0.06954044765896268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,5120,0.0589324434598287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,4096,0.04879022306866116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,4096,0.07114666700363159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,4096,0.046041776736577354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,3584,0.04425955480999417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,3584,0.06383644209967719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,3584,0.041301333241992526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,3072,0.053242666853798755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,3072,0.038607110579808555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,3072,0.036426666710111834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,2560,0.03420533405409919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,2560,0.045490665568245776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,2560,0.03174133433236016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,2048,0.03696088989575704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,2048,0.028928889168633357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,2048,0.02677333354949951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,1536,0.024079999989933435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,1536,0.028797331783506606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,1536,0.02305511136849721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,1024,0.01921688848071628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,1024,0.020977778567208186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,1024,0.018797333041826885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,768,0.01704266667366028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,768,0.01699555582470364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,768,0.01646933290693495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,512,0.014136888914638095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,512,0.013269333375824822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,512,0.014927110738224454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,256,0.008884444832801819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,256,0.011502222054534487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,256,0.012344888514942594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,3072,128,0.010199111368921068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,128,0.007157333195209503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,128,0.011415111521879831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,64,0.006118222243256039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,64,0.011741333537631564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,3072,32,0.0063733334342638654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,3072,32,0.011933333343929715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,65536,0.6165191332499186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,65536,0.929833730061849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,65536,0.6816764407687717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,16384,0.1545519961251153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,16384,0.21814400619930693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,16384,0.16414666175842285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,12288,0.11636088954077826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,12288,0.16632266839345297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,12288,0.12318755520714654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,10240,0.098544889026218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,10240,0.1400444507598877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,10240,0.10378044181399876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,8192,0.11098311344782512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,8192,0.08002844121721056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,8192,0.08423288663228352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,7168,0.09642577833599514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,7168,0.07374133004082574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,6144,0.08477066622840033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,6144,0.06521155436833699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,7168,0.07154133584764269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,6144,0.061907556321885854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,5120,0.07060177458657159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,5120,0.0554657777150472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,4096,0.05776888794369168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,5120,0.05277155505286323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,4096,0.04356888930002848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,4096,0.04485955503251818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,3584,0.05164000060823229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,3584,0.04025955663786994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,3072,0.04427644279268053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,3584,0.03970311085383097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,3072,0.03565866748491923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,3072,0.03449866506788466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,2560,0.030672887961069744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,2560,0.03792622354295518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,2560,0.030943112240897283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,2048,0.026208889153268602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,2048,0.030851556195153132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,2048,0.026576000783178542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,1536,0.021567111214001972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,1536,0.024189333120981853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,1536,0.02247200078434414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,1024,0.01800355480776893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,1024,0.018724444839689467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,768,0.01499466598033905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,768,0.01687733332316081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,1024,0.017543110582563613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,768,0.015092444088723926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,512,0.012783110969596438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,512,0.011681777735551199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,512,0.01460444430510203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,256,0.01054933336046007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,256,0.008201777935028076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,256,0.012149333126015134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2560,128,0.009405333134863112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,128,0.006671999891599019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,128,0.011370666325092316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,64,0.0058311112225055695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,64,0.01179377817445331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2560,32,0.006014222072230444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2560,32,0.011811555259757571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,65536,0.6824737654791938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,65536,0.5198551283942329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,65536,0.5016622013515896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,16384,0.17335466543833414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,16384,0.12793955538007948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,16384,0.12649688455793592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,12288,0.0995111134317186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,12288,0.13118133280012342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,12288,0.09304355250464545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,10240,0.10806666480170356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,10240,0.07908444272147284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,10240,0.08380177948209976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,8192,0.06788177622689141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,8192,0.08996178044213189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,8192,0.06365866793526544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,7168,0.06110044320424398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,7168,0.07762577798631456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,7168,0.0566435522503323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,6144,0.053492443429099194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,6144,0.06752266486485799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,6144,0.04985510971811083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,5120,0.04563644528388977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,5120,0.057251552740732826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,5120,0.04399289025200737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,4096,0.038078222009870745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,4096,0.04602488875389099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,4096,0.03493600090344747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,3584,0.04014577799373203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,3584,0.03173777792188857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,3072,0.030477331744299993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,3072,0.035298665364583336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,3072,0.028404444456100464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,3584,0.034061332543691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,2560,0.030249777767393324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,2560,0.02678844498263465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,2560,0.024872889121373493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,2048,0.022870222727457683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,2048,0.025381333298153345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,2048,0.021743110484547083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,1536,0.019967110620604623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,1536,0.018600889378123812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,1536,0.019018666611777413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,1024,0.01556622154182858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,1024,0.014896889527638754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,1024,0.015505777464972602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,768,0.013456000222100152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,768,0.012467555701732635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,768,0.014226666755146451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,512,0.011784888803958893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,512,0.0098213329911232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,512,0.01181422256761127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,256,0.009909333454238044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,256,0.0070773329999711775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,256,0.010495999621020423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,2048,128,0.008815111385451423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,128,0.005799110978841782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,128,0.009760888914267222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,64,0.00509066672788726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,64,0.010110222631030613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,2048,32,0.0053173332578606075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,2048,32,0.010127999716334874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,65536,0.4142293400234646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,65536,0.5492666562398275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,65536,0.4961182276407878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,16384,0.13463732931349012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,16384,0.1062506635983785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,16384,0.12446488274468316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,12288,0.08105422390831842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,12288,0.10431377755271064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,12288,0.08972977929645115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,10240,0.06906133227878146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,10240,0.08615199724833171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,10240,0.07637777593400744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,8192,0.06788355774349637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,8192,0.056741330358717174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,8192,0.06402844190597534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,7168,0.0503386656443278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,7168,0.0616133345497979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,7168,0.056052446365356445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,6144,0.052505777941809766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,6144,0.04893866512510511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,6144,0.04416444566514757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,5120,0.037885334756639265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,5120,0.04384711053636339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,5120,0.04217511084344652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,4096,0.031882666879230075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,4096,0.03613422314325968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,4096,0.0341288877858056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,3584,0.03190755512979295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,3584,0.031175110075208876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,3584,0.028938667641745672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,3072,0.025582222474945918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,3072,0.02812355425622728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,3072,0.027923554182052612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,2560,0.024057777391539678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,2560,0.022873777482244704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,2560,0.0247999992635515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,2048,0.01980266637272305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,2048,0.019704000817404855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,2048,0.021418665846188862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,1536,0.016531555189026725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,1536,0.01611555616060893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,1536,0.018560000591807894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,1024,0.013902222116788229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,1024,0.012114666402339935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,1024,0.01555022266176012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,768,0.010249777800507015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,768,0.014149333039919535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,512,0.010753778119881948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,512,0.00777511133088006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,768,0.011980444192886353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,512,0.011992000043392181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,256,0.008841777841250101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,256,0.005974222388532426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,256,0.010506666368908353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1536,128,0.008047111332416534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,128,0.005042666776312722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,128,0.009878222313192155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,64,0.004470222112205294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,64,0.010082667072614035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1536,32,0.0046977777447965415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1536,32,0.010124444133705562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,65536,0.3331955538855659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,65536,0.36375821961296934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,65536,0.3466595543755426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,16384,0.08732710944281684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,16384,0.09208089113235474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,16384,0.08498222298092312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,12288,0.07043288813696967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,12288,0.06688355737262301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,12288,0.061908443768819175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,10240,0.05671111080381605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,10240,0.059339556429121226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,10240,0.053751998477511935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,8192,0.04773688978619046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,8192,0.04533511069085863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,7168,0.04129511117935181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,8192,0.04660355713632372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,7168,0.041466666592492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,7168,0.04013955593109131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,6144,0.036582221587498985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,6144,0.036369777388042875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,6144,0.035192890299691096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,5120,0.030404445197847154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,5120,0.0314337776766883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,4096,0.025443555580245122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,5120,0.03138577938079834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,4096,0.026337777574857075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,4096,0.024650666448805068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,3584,0.02232444451914893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,3584,0.02328088879585266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,3584,0.024308444725142583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,3072,0.019736889335844252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,3072,0.02117066747612423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,3072,0.021436444587177698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,2560,0.01703377730316586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,2560,0.019039111004935372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,2560,0.019391111201710172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,2048,0.014304889572991265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,2048,0.01680355601840549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,2048,0.016749334004190233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,1536,0.011752888560295105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,1536,0.014263111684057446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,1536,0.014969777729776172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,1024,0.009177777502271865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,1024,0.0120319997270902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,1024,0.012886222038004132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,768,0.010316444767846001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,768,0.007721777591440413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,768,0.011973333027627734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,512,0.008451555338170793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,512,0.006145777801672618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,512,0.009840000006887648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,256,0.004964444372389051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,256,0.0069857777820693115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,256,0.00908622228437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,1024,128,0.006228444476922353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,128,0.004306666553020477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,128,0.008445333275530074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,64,0.003850666599141227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,64,0.00849866701496972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,1024,32,0.004096888833575779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,1024,32,0.008638222184446123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,65536,0.3212960031297472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,65536,0.28428710831536186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,65536,0.32433509826660156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,16384,0.08427021900812785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,16384,0.07723644706938002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,16384,0.0760257773929172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,12288,0.06485955582724677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,12288,0.05548000004556444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,12288,0.05791022380193075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,10240,0.055339554945627846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,10240,0.045612444480260216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,10240,0.05026222268740336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,8192,0.03755377729733785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,8192,0.04583200150065952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,8192,0.04110488957828946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,7168,0.03378044565518697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,7168,0.037271999650531344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,6144,0.028726223442289565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,7168,0.040644443697399564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,6144,0.03364355696572198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,6144,0.03561866614553664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,5120,0.024350222614076402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,5120,0.030925333499908447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,5120,0.029476443926493328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,4096,0.025898666845427618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,4096,0.0199617776605818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,4096,0.023736000061035156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,3584,0.01772533357143402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,3584,0.021990221407678392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,3072,0.021067554752031963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,3072,0.015840889679061044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,3072,0.020133333073721994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,3584,0.023792889383104112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,2560,0.013566222455766467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,2560,0.018426666657129925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,2560,0.019107555349667866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,2048,0.011716444459226398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,2048,0.016420443852742512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,2048,0.016562667157914903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,1536,0.009701333112186855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,1536,0.014666666587193808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,1536,0.01400888959566752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,1024,0.007804444266690149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,1024,0.012568000290128918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,1024,0.011934222446547614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,768,0.006595555692911148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,768,0.009644444617960189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,768,0.011114666859308878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,512,0.005333333379692501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,512,0.009742221898502773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,512,0.007829333345095316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,256,0.004504888835880491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,256,0.008971555365456475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,128,0.003911111089918348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,128,0.008347555167145198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,256,0.006441777778996362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,768,128,0.005850666513045629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,64,0.0035662220584021676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,64,0.008454221818182204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,768,32,0.0037768888804647657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,768,32,0.00852711084816191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,65536,0.27113244268629283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,65536,0.19007199340396455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,65536,0.3117404513888889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,16384,0.05360355642106798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,16384,0.07156888643900554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,16384,0.07534755600823297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,12288,0.055043554968304105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,12288,0.04198222359021505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,12288,0.057645334137810596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,10240,0.03346488873163859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,10240,0.049921777513292104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,8192,0.03862133291032579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,8192,0.02678666677739885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,10240,0.046853333711624146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,8192,0.04090044564670987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,7168,0.03464444478352865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,7168,0.023936000135209825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,7168,0.03702044486999512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,6144,0.021007110675175984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,6144,0.033415112230512835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,6144,0.030347555875778198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,5120,0.02648266653219859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,5120,0.018191110756662156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,5120,0.029189334975348577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,4096,0.01552177800072564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,4096,0.02233244478702545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,4096,0.02348800003528595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,3584,0.013800000150998434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,3584,0.02163999941613939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,3072,0.011892444557613797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,3584,0.020851555797788832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,3072,0.020068445139461093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,3072,0.018422222799725003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,2560,0.010482666393121084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,2560,0.016689777374267578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,2560,0.018240888913472492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,2048,0.014280888769361707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,2048,0.009179555707507664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,2048,0.016207999653286405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,1536,0.012359999948077731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,1536,0.00810933361450831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,1536,0.014511111709806653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,1024,0.006198222024573221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,1024,0.012162666353914471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,1024,0.010045333041085137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,768,0.008513777620262569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,768,0.005466666486528184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,768,0.010649777948856354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,512,0.0046942221621672315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,512,0.009461333354314169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,512,0.006991110742092133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,256,0.005997333261701796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,256,0.004033777862787247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,256,0.008813333180215623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,512,128,0.005397333453098933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,128,0.0036480000449551474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,128,0.008224888808197444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,64,0.0033866665843460294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,64,0.008329778081840938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,512,32,0.003516444315512975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,512,32,0.008344000412358178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,65536,0.2374053266313341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,65536,0.1078346636560228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,65536,0.3110995557573107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,16384,0.03211022085613675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,16384,0.06434666448169284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,16384,0.07562044594022962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,12288,0.028212444649802312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,12288,0.05739555756251017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,12288,0.048298666874567665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,10240,0.04182933436499702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,10240,0.024476443727811176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,10240,0.04940711127387153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,8192,0.022703111171722412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,8192,0.04046577877468533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,7168,0.01974933346112569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,7168,0.03657599952485826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,6144,0.01457155578666263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,8192,0.03457244568400913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,7168,0.030505776405334473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,6144,0.026687999566396076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,6144,0.03250577714708116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,5120,0.013016888664828407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,5120,0.02883022361331516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,4096,0.019606222709019978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,4096,0.011819555527634092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,4096,0.023258666197458904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,5120,0.02344888945420583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,3584,0.01841600073708428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,3584,0.010563555690977307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,3584,0.021589333812395733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,3072,0.009026666482289633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,3072,0.020080889264742535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,3072,0.01647111111217075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,2560,0.00814844419558843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,2560,0.01811733345190684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,2560,0.014918223023414612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,2048,0.012960889273219638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,2048,0.007160888777838812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,2048,0.016032889485359192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,1536,0.006304889089531369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,1536,0.01089244418674045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,1536,0.01439377831088172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,1024,0.008621333373917473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,1024,0.005032000028424793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,1024,0.01162755572133594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,768,0.004507555729813046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,768,0.010303999814722273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,512,0.004016889052258598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,512,0.009283555878533257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,768,0.007380444142553542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,512,0.0064239998658498125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,256,0.003617777592606015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,256,0.008615111311276754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,128,0.005103999955786599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,256,256,0.005569777968857024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,128,0.0032702222880389956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,128,0.008035555481910706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,64,0.0030782222747802734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,64,0.00814488861295912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,256,32,0.0032924444725116095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,256,32,0.008197333249780867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,65536,0.07974666357040405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,65536,0.31087199846903485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,65536,0.21551288498772514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,16384,0.02591288917594486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,16384,0.05797244442833794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,16384,0.07534044318728976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,12288,0.04502133197254605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,12288,0.020795555578337777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,12288,0.05721866422229343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,10240,0.018559111489189994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,10240,0.04950933323966133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,10240,0.03840444485346476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,8192,0.01682933337158627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,8192,0.040241777896881104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,8192,0.03159288896454705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,7168,0.02809422214825948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,7168,0.014690667390823364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,7168,0.036369777388042875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,6144,0.013402666482660504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,6144,0.024681778417693243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,6144,0.03227555420663621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,5120,0.021546666820844013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,5120,0.01239911135700014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,5120,0.028720001379648846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,4096,0.018432888719770644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,4096,0.011247110863526663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,4096,0.023175110419591267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,3584,0.016726222303178575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,3584,0.010212444596820408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,3584,0.021567111214001972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,3072,0.015063999427689446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,3072,0.009043555292818282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,3072,0.019911110401153564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,2560,0.013758222262064615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,2560,0.008020444048775567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,2560,0.018048000004556444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,2048,0.007259555160999298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,2048,0.015996444556448195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,1536,0.010101333260536194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,1536,0.005927111125654644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,2048,0.011986666255527072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,1536,0.014464888307783337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,1024,0.004834666848182678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,1024,0.01145600030819575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,768,0.004344888859324985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,1024,0.007892444729804993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,768,0.010114666488435533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,768,0.007063111497296228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,512,0.003826666623353958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,512,0.006139555739031897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,512,0.009155555731720394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,256,0.00535111129283905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,256,0.0033866665843460294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,256,0.008543999658690559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,128,0.00309599998096625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,128,0.008034666379292807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2048,128,128,0.004927111168702443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,64,0.0029599999801980127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,64,0.008099555969238281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,128,32,0.003003555453485913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,128,32,0.008114666574531132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,65536,0.07289066579606798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,16384,0.023265777362717524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,65536,0.31108887990315753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,16384,0.07581155829959445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,12288,0.019010666343900893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,12288,0.05692178010940552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,10240,0.016985777351591323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,10240,0.049160000350740224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,8192,0.015033778217103748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,8192,0.040263109736972384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,7168,0.013688888814714221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,7168,0.03631555702951219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,6144,0.012576888832781049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,6144,0.03232889042960273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,5120,0.011562666959232755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,5120,0.028601778878106013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,4096,0.010462221999963125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,4096,0.023169777459568445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,3584,0.009663999908500248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,3584,0.021597334080272253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,3072,0.009170666337013245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,3072,0.019730667273203533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,2560,0.0075493330756823225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,2560,0.017840888765123155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,2048,0.006799110935793982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,2048,0.015962666935390897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,1536,0.005588444570700328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,1536,0.014214222629865011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,1024,0.004320888883537716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,1024,0.011218667030334473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,768,0.003987555702527364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,768,0.010093332992659675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,512,0.003605333467324575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,512,0.009149333669079674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,256,0.0032622222271230486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,256,0.008504889077610439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,128,0.002943111169669363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,128,0.008007999923494127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,64,0.002834666727317704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,64,0.008027555214034187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,64,32,0.0029013332807355454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,64,32,0.008045333127180735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,65536,0.07527466615041097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,16384,0.023595554961098567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,16384,0.07584710915883382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,65536,0.3109297752380371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,12288,0.019088000059127808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,12288,0.05672888623343574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,10240,0.01682222220632765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,10240,0.04948444498909844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,8192,0.01440711153878106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,8192,0.04010844561788771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,7168,0.013361777696344586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,7168,0.03627289003796048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,6144,0.01237155579858356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,6144,0.03222489025857713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,5120,0.011207111179828644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,5120,0.02870044443342421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,4096,0.010073777702119615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,4096,0.023179555932680767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,3584,0.009303111169073317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,3584,0.02146044373512268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,3072,0.008847999903890822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,3072,0.019704888264338177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,2560,0.007244444555706448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,2560,0.01780177818404304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,2048,0.006634666687912411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,2048,0.01591377788119846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,1536,0.00573333352804184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,1536,0.014157333307796054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,1024,0.004293333325121137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,1024,0.011314666933483548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,768,0.003949333396222857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,768,0.010113777385817634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,512,0.003647110942337248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,512,0.009183111290136972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,256,0.0032346666687064697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,256,0.00851288851764467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,128,0.002965333354141977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,128,0.007996444072988298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,64,0.0027662221756246355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,64,0.008047111332416534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2048,32,32,0.002975111206372579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2048,32,32,0.008053333395057255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,16384,1.7088791529337566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,12288,1.1972373326619465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,16384,3.4332586924235025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,16384,1.6359715991550023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,12288,2.5436070760091147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,12288,1.172027587890625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,10240,1.0357146792941623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,10240,0.9649510913425021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,10240,2.0960676405164933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,8192,0.7661031087239584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,8192,1.5882701873779297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,8192,0.7253866725497775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,7168,0.6690115398830838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,7168,0.6904053158230252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,7168,1.4443288379245336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,6144,1.2351804309421115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,6144,0.5781920221116807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,6144,0.5527333153618706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,5120,0.4562364684210883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,5120,1.0074613359239366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,5120,0.4898106786939833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,4096,0.3940097755855984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,4096,0.8094853295220269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,4096,0.37996355692545575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,3584,0.3481706778208415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,3584,0.6902924643622504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,3584,0.3224293390909831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,3072,0.2961617840660943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,3072,0.5684551133049859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,3072,0.2847964498731825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,2560,0.4885848893059625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,2560,0.25884887907240123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,2560,0.2395084434085422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,2048,0.2104560004340278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,2048,0.38147465387980145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,2048,0.19150400161743164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,1536,0.16323999563852945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,1536,0.2952408790588379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,1536,0.1558613379796346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,1024,0.1189146704143948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,1024,0.2027111053466797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,1024,0.11731199423472087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,768,0.09456000063154434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,768,0.15698311063978407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,768,0.0985777775446574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,512,0.07187555233637492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,512,0.1144755548901028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,512,0.08531110816531712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,256,0.06849422057469685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,256,0.07168800301022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,128,0.051855110459857516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,128,0.05712622404098511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,128,0.0673182209332784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,65536,256,0.05519377854135302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,64,0.049434665176603526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,64,0.06838666730456881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,65536,32,0.0491884450117747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,65536,32,0.0694284439086914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,16384,0.3916800022125244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,65536,1.738483534918891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,16384,0.7571013238694934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,65536,1.7139413621690538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,65536,3.2902889251708984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,16384,0.3881066640218099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,12288,0.29548801316155326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,12288,0.2868559890323215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,12288,0.5310346815321181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,10240,0.2441448900434706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,10240,0.45876444710625547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,10240,0.2399697833591037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,8192,0.2022693289650811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,8192,0.36758489078945583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,8192,0.19054310851626924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,7168,0.17579289277394614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,7168,0.3122062153286404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,7168,0.164865771929423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,6144,0.27473778194851345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,6144,0.1543253395292494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,6144,0.14389333460066053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,5120,0.12978222635057238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,5120,0.22965865665011934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,5120,0.12294933531019424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,4096,0.10359466738171047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,4096,0.18252266777886283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,4096,0.09934310780631171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,3584,0.16229866610633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,3584,0.08764444457160102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,3072,0.14291644096374512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,3584,0.09452444314956665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,3072,0.0804435544543796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,3072,0.07495911253823175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,2560,0.11971910794576009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,2560,0.06853510936101277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,2560,0.06535110870997111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,2048,0.09815111425187852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,2048,0.05418222149213155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,1536,0.045816888411839805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,1536,0.07349689139260186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,1536,0.04271466533342997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,1024,0.03427822060055203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,2048,0.05766933494144016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,1024,0.05275377631187439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,1024,0.033942222595214844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,768,0.04231999980078804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,768,0.028730667299694482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,768,0.030340443054835003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,512,0.02314311100376977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,512,0.03198933270242479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,512,0.02697155541843838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,256,0.02184533410602146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,256,0.022868444522221882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,128,0.016723554995324876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,256,0.018244443668259513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,16384,128,0.01567911108334859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,128,0.021351110604074266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,64,0.013728000223636627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,64,0.0210844443904029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,16384,32,0.01426577733622657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,16384,32,0.021236444513003033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,65536,2.4470365312364364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,65536,1.3166578080919054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,65536,1.1857582728068035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,16384,0.5269635518391927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,16384,0.2893582185109456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,12288,0.22739911079406738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,12288,0.4051155514187283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,12288,0.2181288931104872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,16384,0.29853065808614093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,10240,0.18587999873691133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,10240,0.17880800035264757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,10240,0.3425973256429036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,8192,0.15290666951073542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,8192,0.2739582326677111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,8192,0.14625244670444065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,7168,0.13407288657294378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,7168,0.23920000924004448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,7168,0.1260062191221449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,6144,0.11468444930182563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,6144,0.20861511760287813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,6144,0.11151556173960368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,5120,0.1718604432211982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,5120,0.09347644117143418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,5120,0.09979022211498684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,4096,0.13697689109378391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,4096,0.07570222351286146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,4096,0.08073955774307251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,3584,0.07268266545401679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,3584,0.12306666374206543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,3584,0.06782666842142741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,3072,0.1079359981748793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,3072,0.06261333492067125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,3072,0.058641778098212353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,2560,0.0907484425438775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,2560,0.05338399940066867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,2560,0.0513591104083591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,2048,0.0731946693526374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,2048,0.04231199953291151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,1536,0.05594755543602837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,1536,0.03386222322781881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,2048,0.045532445112864174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,1024,0.027668444646729365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,1024,0.04115555683771769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,1024,0.027089777919981215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,1536,0.03651911020278931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,768,0.03308799862861633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,768,0.02427111069361369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,512,0.024975111087163288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,512,0.021346666746669348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,768,0.023387556274731953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,256,0.015035554766654968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,512,0.019448000523779128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,256,0.017504000001483493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,256,0.017998221847746108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,12288,128,0.013195555243227216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,128,0.013801777528391944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,128,0.01555288831392924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,64,0.011370666325092316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,64,0.01626399987273746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,12288,32,0.011521778172916837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,12288,32,0.016872000363137987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,65536,1.2929599550035265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,16384,0.30197156800164116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,16384,0.4554168913099501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,65536,0.9894382688734266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,65536,2.0796356201171875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,16384,0.24357244703504774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,12288,0.18325689103868273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,12288,0.33388532532585996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,12288,0.2242586612701416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,10240,0.2865991062588162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,10240,0.15159911579555935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,8192,0.22808265686035156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,8192,0.12247288227081299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,10240,0.191264894273546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,8192,0.15281689167022705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,7168,0.20000978310902914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,7168,0.13342132833268908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,7168,0.11009244124094646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,6144,0.11526578002505833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,6144,0.17258666621314156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,6144,0.09372178051206802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,5120,0.09957689046859741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,5120,0.1452817784415351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,5120,0.0800515545739068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,4096,0.1182017723719279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,4096,0.08059377802742852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,4096,0.06445333030488756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,3584,0.07323822047975329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,3584,0.1023626658651564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,3584,0.058861335118611656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,3072,0.06285599867502849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,3072,0.08902133173412746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,3072,0.0510391096274058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,2560,0.05542133251825968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,2560,0.07406400309668647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,2560,0.043303999635908336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,2048,0.04641422298219469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,2048,0.06237955225838555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,2048,0.03612800108061896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,1536,0.036848889456854925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,1536,0.04810488886303372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,1536,0.02959377898110284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,1024,0.03346488873163859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,1024,0.029846221208572388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,1024,0.024041778511471216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,768,0.02682133350107405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,768,0.024578667349285547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,768,0.021260443660948012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,512,0.020879111356205408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,512,0.02018222212791443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,512,0.018754666050275166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,256,0.015840889679061044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,256,0.013978666729397245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,256,0.015637333194414776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,128,0.010606221854686737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,128,0.013564444250530668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,64,0.008700444466537898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,64,0.014552889598740472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,10240,32,0.009047110875447592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,10240,32,0.015127999915017022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,65536,0.8432462480333117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,65536,1.5422275331285265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,10240,128,0.013870221873124441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,16384,0.35913244883219403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,16384,0.19535999827914766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,16384,0.20238044526841906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,65536,0.848264906141493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,12288,0.27208267317877877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,12288,0.1549546718597412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,12288,0.15174133247799343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,10240,0.1328844494289822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,10240,0.2243582142723931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,10240,0.12388888994852702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,8192,0.10534133513768514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,8192,0.18100533220503065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,8192,0.10194133387671576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,7168,0.15885867012871638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,7168,0.0905768871307373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,6144,0.1345591147740682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,6144,0.07669599850972493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,6144,0.08093955781724718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,7168,0.09369688563876682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,5120,0.11593333880106609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,5120,0.06574310859044392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,5120,0.06926133235295613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,4096,0.09140088823106553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,4096,0.05728889173931546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,4096,0.05341688791910807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,3584,0.08141866657469007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,3584,0.051274667183558144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,3584,0.048135999176237315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,3072,0.07099911239412096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,3072,0.04136355717976888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,3072,0.04449422160784403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,2560,0.03886222177081638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,2560,0.058840003278520375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,2560,0.03638933433426751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,2048,0.04936266607708401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,2048,0.03013955553372701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,1536,0.02684088879161411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,1536,0.03773244553142124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,1536,0.024689777029885188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,1024,0.021330666210916307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,1024,0.02684088879161411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,2048,0.03305866652064853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,1024,0.020438222421540152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,768,0.018360000517633226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,768,0.02149066660139296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,768,0.018192888961897958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,512,0.015407111909654407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,512,0.016528889536857605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,512,0.01594311164485084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,256,0.012346666720178394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,256,0.011739555332395764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,256,0.012873777912722694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,8192,128,0.01090666651725769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,128,0.008836444053384993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,128,0.011794666449228922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,64,0.007304000357786815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,64,0.01258488910065757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,8192,32,0.0076657773719893555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,8192,32,0.012810666528013019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,16384,0.20987466971079508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,16384,0.3314559989505344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,65536,0.785554673936632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,65536,1.480308426751031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,16384,0.19223911232418486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,65536,0.8332133293151855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,12288,0.15665066242218018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,12288,0.25700444645351833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,12288,0.14686755339304605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,10240,0.20916444725460479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,10240,0.1309244500266181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,10240,0.1240382194519043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,8192,0.10669777790705363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,8192,0.1713822285334269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,8192,0.09982844193776448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,7168,0.1527715524037679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,7168,0.08660444286134507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,6144,0.08186044295628865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,7168,0.09526755412419637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,6144,0.12970044877794054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,6144,0.07585244708591037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,5120,0.10845600234137641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,5120,0.07004533211390178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,5120,0.06444533003701104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,4096,0.08644177516301473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,4096,0.052857776482899986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,3584,0.05270399981074863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,4096,0.05745955308278402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,3584,0.07716888851589628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,3584,0.047154668304655284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,3072,0.045356445842319064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,3072,0.06832533412509494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,3072,0.04137688875198364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,2560,0.04049777653482225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,2560,0.05704266495174832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,2560,0.03537777728504605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,2048,0.0340924428568946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,2048,0.04626133375697666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,2048,0.02972888946533203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,1536,0.02757511039574941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,1536,0.03590844406021966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,1536,0.024712888730896845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,1024,0.025778666138648987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,1024,0.02050755586889055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,1024,0.02240266733699375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,768,0.02100177771515316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,768,0.018352000249756705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,768,0.018996444013383653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,512,0.016311999824311998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,512,0.0162808895111084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,512,0.016007999579111736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,256,0.011349333657158745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,256,0.012888888518015543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,256,0.012587555580668978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,7168,128,0.011576889289749993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,128,0.008252444366614023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,128,0.01186044431395001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,64,0.007116444408893585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,64,0.01256711118751102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,7168,32,0.007676444119877285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,7168,32,0.012664000193277994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,65536,1.204051547580295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,65536,0.6541066699557835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,16384,0.15652533372243246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,65536,0.6347875595092773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,16384,0.2806871202256945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,16384,0.15092178185780844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,12288,0.12092355887095134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,12288,0.20952533351050484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,12288,0.1191271146138509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,10240,0.10112266408072577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,10240,0.17180000411139595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,10240,0.09700800312889947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,8192,0.1383662223815918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,8192,0.07891377475526598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,8192,0.08335644669002956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,7168,0.07303555806477864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,7168,0.1258666647805108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,7168,0.06975111034181383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,6144,0.06320355335871379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,6144,0.10656266742282444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,6144,0.060030221939086914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,5120,0.05406755540106031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,5120,0.09031199746661717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,5120,0.052074667480256825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,4096,0.045055111249287925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,4096,0.07175288597742717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,4096,0.04271999994913737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,3584,0.04056355688307021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,3584,0.06290133131874932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,3584,0.038146668010287814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,3072,0.05510044429037306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,3072,0.033731556600994535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,2560,0.031607998741997614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,3072,0.0356026656097836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,2560,0.045666668150160045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,2560,0.029474665721257527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,2048,0.02696177860101064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,2048,0.03784088955985175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,2048,0.024978665841950312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,1536,0.022409778502252366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,1536,0.0292524430486891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,1536,0.021108445194032457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,1024,0.018034666776657104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,1024,0.02104533380932278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,1024,0.017335999343130324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,768,0.0159582214223014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,768,0.017101332545280457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,768,0.015559110376569959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,512,0.013413333230548434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,512,0.013774221969975365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,512,0.013110222087966071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,256,0.009353777600659264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,256,0.011063999599880643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,128,0.010011555420027839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,128,0.007127111156781514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,128,0.010230222509966956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,6144,256,0.011418666276666852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,64,0.006164444403515921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,64,0.010868444210953183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,6144,32,0.006457777900828256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,6144,32,0.011159111228254108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,65536,0.9092044830322266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,65536,0.6076986524793837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,16384,0.22130844328138563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,65536,0.5673493279351128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,16384,0.1472488906648424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,12288,0.10427644517686631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,16384,0.13881422413720026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,12288,0.16758044560750326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,12288,0.1135866641998291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,10240,0.0899582240316603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,10240,0.13985599411858454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,10240,0.09504355324639215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,8192,0.07151822249094646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,8192,0.11141688293880886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,8192,0.07676888836754693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,7168,0.10101688570446438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,7168,0.06392711400985718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,7168,0.06821778085496691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,6144,0.05578577518463135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,6144,0.0860533316930135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,6144,0.059134225050608315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,5120,0.07199733124838935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,5120,0.050382223394181996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,4096,0.039368887742360435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,5120,0.04713333315319485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,4096,0.05926577912436592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,4096,0.041702220837275185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,3584,0.05101422137684292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,3584,0.03724266754256354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,3584,0.03597155544492934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,3072,0.03161599900987413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,3072,0.044927110274632774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,3072,0.03260088960329691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,2560,0.038093331787321306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,2560,0.02888710962401496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,2560,0.028177777926127117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,2048,0.024285333024130926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,2048,0.03139288889037238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,2048,0.024295111497243244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,1536,0.024289778537220422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,1536,0.020036444067955017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,1536,0.020776889390415616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,1024,0.016207110550668504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,1024,0.01825777855184343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,1024,0.01719288859102461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,768,0.014945778581831189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,768,0.01551466683546702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,512,0.011824000212881299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,512,0.012598222328556908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,768,0.01456622282663981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,512,0.012435555458068848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,256,0.010527999864684211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,256,0.008339555727110969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,256,0.010976889067225985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,128,0.006776888751321369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,128,0.010194666683673859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,5120,128,0.009356444080670675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,64,0.0058782220714622075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,64,0.010772444307804108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,5120,32,0.006062222023804982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,5120,32,0.010983111129866706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,65536,0.44537512461344403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,65536,0.7412266731262207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,65536,0.42998843722873265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,16384,0.11285511652628581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,16384,0.17661689387427437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,16384,0.10822488864262898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,12288,0.08715644147660996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,12288,0.13312710656060114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,12288,0.08339999781714545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,10240,0.07368266582489014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,10240,0.11007289091746013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,10240,0.0698488884501987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,8192,0.059826665454440646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,8192,0.09007110860612656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,8192,0.057201776239607066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,7168,0.053116444084379405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,7168,0.08030844397015043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,7168,0.0510062211089664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,6144,0.06712266471650866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,6144,0.04421866602367825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,6144,0.04592000113593208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,5120,0.05826222234302097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,5120,0.04003644320699904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,5120,0.038132445679770574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,4096,0.04647466540336609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,4096,0.03173689047495524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,3584,0.04065777858098348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,3584,0.028825776444541082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,3072,0.03591555688116286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,3072,0.025847110483381484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,2560,0.024311999479929607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,4096,0.03363733490308126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,3584,0.03071999881002638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,3072,0.02734311090575324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,2560,0.030138668086793687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,2560,0.02274666726589203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,2048,0.025432000557581585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,2048,0.0196115556690428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,2048,0.021106666988796655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,1536,0.020040000478426617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,1536,0.016695999436908297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,1536,0.01755288905567593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,1024,0.014633778068754407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,1024,0.01517777807182736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,1024,0.013870221873124441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,768,0.013087111214796701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,768,0.01255022237698237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,768,0.01198755535814497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,512,0.010069333016872406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,512,0.01144977741771274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,512,0.010552888943089379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,256,0.00980888886584176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,256,0.00720711135201984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,256,0.009168888959619734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,128,0.005882666756709416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,4096,128,0.008749333520730337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,128,0.00867555538813273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,64,0.005161777966552311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,64,0.00922400007645289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,4096,32,0.005375111268626318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,4096,32,0.009335999687512716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,65536,0.651168876224094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,65536,0.4382488992479112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,65536,0.4219244321187337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,16384,0.16466311613718668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,16384,0.1049013336499532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,16384,0.1149937841627333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,12288,0.08776622348361546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,12288,0.1237679984834459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,12288,0.08003644148508708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,10240,0.10483555661307441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,10240,0.06776977909935845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,10240,0.07406755288441975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,8192,0.08557333548863728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,8192,0.05523022347026401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,8192,0.06002666552861532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,7168,0.07449244128333198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,7168,0.049154665735032826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,6144,0.0638186666700575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,6144,0.04295644495222303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,7168,0.05442666676309374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,6144,0.04710666669739617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,5120,0.04093777802255418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,5120,0.053341332409116954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,5120,0.0370488895310296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,4096,0.0440942214594947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,4096,0.031015111340416804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,4096,0.03425155414475335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,3584,0.03179733289612664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,3584,0.038995554049809776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,3584,0.02831733226776123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,3072,0.03422044383154975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,3072,0.025288888149791296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,3072,0.027671999401516382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,2560,0.0247697780529658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,2560,0.029383109675513372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,2560,0.0225591113169988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,2048,0.024554666545655992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,2048,0.0214720004134708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,2048,0.019521777828534443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,1536,0.019482667247454327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,1536,0.017942221628295053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,1536,0.016484444340070088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,1024,0.015030221806632148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,1024,0.013780444032616086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,1024,0.015136000182893542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,768,0.013283555706342062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,768,0.012431110772821637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,768,0.01202933324707879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,512,0.011711111499203576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,512,0.009721777505344814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,512,0.010221333139472539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,256,0.0070222218831380205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,256,0.009248889154858058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,256,0.009947555760542551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,128,0.0058542220956749385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3584,128,0.00889511075284746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,128,0.008624888956546783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,64,0.005074666606055366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,64,0.009082666701740688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3584,32,0.005401777724424998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3584,32,0.009280888570679558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,65536,0.556679990556505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,65536,0.40863998730977374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,65536,0.35103999243842227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,16384,0.08940088748931885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,16384,0.13769243823157415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,16384,0.10449600219726562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,12288,0.10410310824712117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,12288,0.07925511068767972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,12288,0.06909422079722087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,10240,0.08870044681761001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,10240,0.06656710969077216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,10240,0.05818133221732246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,8192,0.04772355490260654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,8192,0.07102933194902208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,8192,0.05429688758320279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,7168,0.04227377639876472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,7168,0.06134044461780124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,7168,0.04821955495410495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,6144,0.05327466792530484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,6144,0.042592889732784696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,5120,0.04502400093608432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,6144,0.037618666887283325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,5120,0.03292444348335266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,5120,0.036659555302725896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,4096,0.03703555464744568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,4096,0.03061244553989834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,3584,0.03230666783120897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,3584,0.02780888809098138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,3072,0.027850667635599773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,3072,0.024877332978778418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,2560,0.023931556277804907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,2560,0.022219555245505437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,4096,0.02789155642191569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,3584,0.02535377773973677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,2048,0.01998133295112186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,2048,0.01928711103068458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,1536,0.016122667325867545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,1536,0.01647911138004727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,1024,0.012136889000733694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,3072,0.022459555003378127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,1024,0.01367911116944419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,2560,0.020106666617923312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,2048,0.017753778232468497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,1536,0.014905777242448596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,1024,0.012662221988042196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,768,0.010219555762079027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,768,0.011500444677140979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,512,0.0103511114915212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,512,0.008249777886602614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,512,0.010240000155236986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,256,0.00599911105301645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,256,0.00925155563486947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,768,0.011608888705571493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,256,0.00885333369175593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,128,0.005046222358942032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,3072,128,0.007830222447713217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,128,0.008608000146018134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,64,0.004662222332424588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,64,0.00906755526860555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,3072,32,0.0046737777690092725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,3072,32,0.009305777649084726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,65536,0.4526835547553168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,65536,0.41045334604051376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,65536,0.30019111103481716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,16384,0.07938310835096571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,16384,0.12501688798268637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,16384,0.10428533289167617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,12288,0.08886044555240208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,12288,0.07880977789560954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,10240,0.07314133644104004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,10240,0.06628355715009901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,8192,0.04238399863243103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,8192,0.0612915555636088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,8192,0.05376977721850077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,7168,0.05221955643759834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,7168,0.048080000612470836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,12288,0.0607831080754598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,10240,0.0522311098045773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,7168,0.03827910953097873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,6144,0.03314933180809021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,6144,0.04665866825315687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,6144,0.04201511210865445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,5120,0.0388337771097819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,5120,0.029306666718588933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,5120,0.03673155440224542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,4096,0.02488888965712653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,4096,0.03177777926127116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,4096,0.030657778183619182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,3584,0.028128001425001357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,3584,0.02772355576356252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,3072,0.02472177810139126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,3584,0.02277066641383701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,3072,0.02020355562369029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,3072,0.025096888343493145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,2560,0.021103999680942957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,2560,0.021987555755509272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,2560,0.018371555540296767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,2048,0.016237333416938782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,2048,0.018037334084510803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,2048,0.01926399932967292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,1536,0.014825777875052558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,1536,0.013620444469981723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,1536,0.01647555496957567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,1024,0.011539555258221097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,1024,0.011549333731333414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,1024,0.012969777815871768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,768,0.010577778021494547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,768,0.009878222313192155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,768,0.011391110718250275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,512,0.007804444266690149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,512,0.009464888936943477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,512,0.010270222193664974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,256,0.007831110722488826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,256,0.0058373332851462895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,256,0.009279111193286048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,128,0.004950222041871813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,128,0.00852711084816191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,64,0.004355555607212914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,64,0.008983111215962304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2560,128,0.006959111326270633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2560,32,0.004559111264016894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2560,32,0.00904977818330129
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,65536,0.2677271101209852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,65536,0.36130399174160427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,65536,0.2559635639190674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,16384,0.069669332769182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,16384,0.09133866760465835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,16384,0.06491200129191081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,10240,0.04318399892912971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,12288,0.05371110969119602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,12288,0.06942133108774821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,12288,0.050491554869545825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,10240,0.05770311090681288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,10240,0.04620000057750278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,8192,0.047783109876844615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,8192,0.03552533189455668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,7168,0.033746666378445096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,7168,0.04157422317398919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,7168,0.03187377916442023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,6144,0.02955555584695604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,6144,0.03623644510904948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,6144,0.02847733431392246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,8192,0.03793333305252923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,5120,0.02627022233274248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,5120,0.030453334252039593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,5120,0.025363556212849084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,4096,0.024608888559871252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,4096,0.02158222264713711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,4096,0.022231111923853557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,3584,0.022218666142887537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,3584,0.019810666640599568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,3072,0.01949155496226417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,3584,0.020596444606781006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,3072,0.018227555685573153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,3072,0.018152889278199937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,2560,0.016618667377365958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,2560,0.016819554898473952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,2560,0.01650044487582313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,2048,0.014845333165592618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,2048,0.014228444960382251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,2048,0.014654222461912366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,1536,0.011718221836619906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,1536,0.012938666674825879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,1536,0.012540444731712341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,1024,0.009195555415418413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,1024,0.010205333431561789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,1024,0.01074577785200543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,768,0.009574222067991892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,768,0.007547555698288812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,768,0.009159999589125315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,512,0.006116444451941385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,512,0.008216000265545314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,512,0.008446222378147973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,256,0.004956444518433677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,256,0.007831999825106727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,128,0.004317333300908406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,128,0.007332444190979004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,64,0.0038453332251972626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,64,0.007607111500369177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,2048,32,0.004023111114899318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,2048,32,0.007662222617202335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,256,0.007035555938879649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,2048,128,0.0061048890153567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,65536,0.27979644139607746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,65536,0.25195466147528756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,65536,0.24221332867940268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,16384,0.06717511018117268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,16384,0.0764568911658393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,16384,0.06167733007007175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,12288,0.05167288912667168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,12288,0.0551475551393297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,12288,0.04801955487993029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,10240,0.04439822170469496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,10240,0.04602488875389099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,10240,0.04125866625044081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,8192,0.03658399979273478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,8192,0.03645155496067471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,8192,0.034272889296213783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,7168,0.0330062210559845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,7168,0.03231466809908549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,7168,0.03084355592727661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,6144,0.029059555795457628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,6144,0.02833066715134515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,6144,0.02754844393995073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,5120,0.025614221890767414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,5120,0.023560000790490046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,5120,0.02474311159716712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,4096,0.021811554829279583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,4096,0.0196435550848643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,4096,0.020849777592553034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,3584,0.020231111182106864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,3584,0.01752800080511305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,3584,0.019447111421161227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,3072,0.015607999430762397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,3072,0.017621333400408428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,2560,0.013599110974205865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,3072,0.01792888840039571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,2560,0.016417778200573392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,2560,0.016176000237464905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,2048,0.014497778481907315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,2048,0.011585777832402123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,2048,0.014440889159838358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,1536,0.009703999592198266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,1536,0.012599999705950419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,1536,0.012098666694429187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,1024,0.010188444621033138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,1024,0.007641777396202087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,1024,0.009951111343171861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,768,0.00868266655339135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,768,0.006364444477690591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,768,0.009157333109113906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,512,0.005315555466545953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,512,0.008319999608728621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,256,0.00444444434510337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,256,0.006229333165619109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,256,0.007723555796676212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,128,0.00566133318675889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,128,0.003970666478077571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,128,0.007271111011505127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,64,0.003576888806290097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,64,0.0074826669361856245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1536,512,0.007244444555706448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1536,32,0.0038008888562520346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1536,32,0.0075599998235702515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,65536,0.20127467314402261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,65536,0.1870115598042806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,65536,0.2322711149851481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,16384,0.0540648897488912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,16384,0.052278220653533936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,16384,0.060956445005204946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,12288,0.04156977931658427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,12288,0.04073866539531284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,12288,0.04737866587109036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,10240,0.032822221517562866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,10240,0.03579911258485582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,10240,0.04039377636379666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,8192,0.026348445150587294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,8192,0.03339733348952399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,8192,0.029885331789652508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,7168,0.02384177843729655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,7168,0.030044443077511255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,6144,0.023874666955735948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,6144,0.0203146669599745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,6144,0.02715555495686001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,5120,0.017465778523021273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,5120,0.024188444018363953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,7168,0.027211555176311072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,4096,0.014692443940374585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,4096,0.020474667350451153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,3584,0.013340444200568728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,3584,0.018830221560266282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,5120,0.021195555726687115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,4096,0.018045332696702745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,3584,0.017200888858901132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,3072,0.015114666687117683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,3072,0.011838222543398539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,3072,0.017391110459963482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,2560,0.010427555276287926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,2560,0.01575288838810391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,2048,0.012338666452301873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,2048,0.009358222285906473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,2048,0.014000000225173103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,2560,0.013916444447305469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,1536,0.007920000288221572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,1536,0.010261333651012844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,1536,0.011928000383906893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,1024,0.008275555239783393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,1024,0.005956444475385878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,1024,0.009464888936943477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,768,0.005387555393907759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,768,0.00888533310757743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,768,0.007296889192528195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,512,0.006478222293986215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,512,0.0047484445903036325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,512,0.0081377774477005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,256,0.004074666649103165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,256,0.007541333635648091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,128,0.0036435553597079385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,128,0.007156444092591603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,64,0.003330666571855545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,64,0.007296000089910295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,1024,32,0.003516444315512975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,1024,32,0.007363555332024892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,65536,0.20338577694363066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,256,0.00572622236278322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,1024,128,0.005326222214433882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,65536,0.14687911669413248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,16384,0.05569689141379463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,16384,0.040758222341537476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,65536,0.2314257886674669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,16384,0.06048711140950521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,12288,0.032335999939176775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,12288,0.046959112087885536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,10240,0.028326223293940227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,10240,0.040221333503723145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,12288,0.04239022069507175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,10240,0.036348445547951594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,8192,0.024847999215126038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,8192,0.033244444264305964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,8192,0.029836446046829224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,7168,0.02677333354949951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,7168,0.0188417782386144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,7168,0.030233777231640283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,6144,0.02366222275627984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,6144,0.016565332810084026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,6144,0.02715822226471371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,5120,0.014632888966136508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,5120,0.021026665965716045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,5120,0.024319999747806128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,4096,0.013896889156765409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,4096,0.020337777005301583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,4096,0.017916444275114272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,3584,0.016688888271649677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,3584,0.011552888486120435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,3584,0.018767111831241183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,3072,0.010050666828950247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,3072,0.017383999294704862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,3072,0.014811555544535318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,2560,0.01350311107105679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,2560,0.008954666554927826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,2560,0.01575999955336253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,2048,0.007827555967701806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,2048,0.011886222494973077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,2048,0.013944889108339945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,1536,0.009415999882751042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,1536,0.006437333507670297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,1536,0.01165422217713462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,1024,0.0052595556610160405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,1024,0.0075297777851422625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,1024,0.009408888717492422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,768,0.0069679998689227635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,768,0.004826666580306159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,768,0.008848889006508721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,512,0.006272888845867581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,512,0.004249777644872665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,128,0.007150222029950883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,512,0.008201777935028076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,256,0.005573333137565189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,256,0.0037191112836201987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,256,0.007589333587222629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,128,0.0034008889148632684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,64,0.0032382222513357797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,64,0.007317333585686154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,768,32,0.0033004445334275565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,768,128,0.005140444470776452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,768,32,0.007351111206743453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,65536,0.10100444157918294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,65536,0.23280710644192168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,65536,0.16183466381496853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,16384,0.02996888756752014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,16384,0.06062844726774427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,16384,0.04457688993877835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,12288,0.024716445141368445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,12288,0.04619466596179538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,12288,0.03383377856678433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,10240,0.029482665989134047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,10240,0.02166755497455597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,10240,0.03974755605061849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,8192,0.0188417782386144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,8192,0.024837333295080397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,8192,0.03316444489690993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,7168,0.022258667482270136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,7168,0.017079111602571275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,7168,0.0302800006336636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,6144,0.019682667321628995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,6144,0.013906665974193148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,6144,0.02703199949529436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,5120,0.012114666402339935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,5120,0.023853333459960088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,4096,0.015237333046065437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,4096,0.012108444339699216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,4096,0.02019466625319587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,3584,0.014465777410401238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,5120,0.017705778280893963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,3584,0.00963288876745436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,3584,0.018615111708641052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,3072,0.00871377769443724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,3072,0.01717422240310245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,2560,0.008090666598743862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,2560,0.015555555621782938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,3072,0.012764444781674279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,2560,0.011512888802422417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,2048,0.010095111197895473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,2048,0.007051555646790399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,2048,0.01366311146153344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,1536,0.00601244428091579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,1536,0.011323555476135679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,1024,0.004872888740566042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,1024,0.009319999979601966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,1536,0.008435555630260045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,768,0.004507555729813046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,768,0.008786666724416945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,512,0.0059075554211934405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,512,0.004048000193304486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,512,0.00812888890504837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,1024,0.0070497774415545994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,256,0.003564444267087512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,256,0.0074720001882976955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,768,0.006504000061088138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,128,0.0033048888047536216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,128,0.007106666763623555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,64,0.003091555502679613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,64,0.007223111059930589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,512,32,0.0032524443748924467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,512,32,0.00721955547730128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,256,0.005294222384691238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,512,128,0.004960888789759742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,65536,0.0565848880343967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,65536,0.23161421881781685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,65536,0.14634577433268228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,16384,0.021436444587177698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,16384,0.06026311053170098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,12288,0.01689511040846507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,12288,0.0462026662296719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,10240,0.015097777048746744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,16384,0.0402471125125885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,10240,0.039670222335391574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,8192,0.02269511090384589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,8192,0.015092444088723926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,12288,0.031347556246651545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,10240,0.027048889133665297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,8192,0.03325066632694668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,7168,0.013600889179441663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,7168,0.030082666211658057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,7168,0.020438222421540152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,6144,0.01829333272245195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,6144,0.0121760004096561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,6144,0.02683911058637831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,5120,0.011332444846630096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,5120,0.02380977736579047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,5120,0.01643111142847273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,4096,0.01038755559258991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,4096,0.014111111561457316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,4096,0.020241777102152508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,3584,0.013321778012646569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,3584,0.009100444614887238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,3584,0.018518222702874076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,3072,0.008078222473462423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,3072,0.016939555605252583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,2560,0.010286221901575724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,2560,0.007837333612971837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,3072,0.011751111182901593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,2560,0.015414221419228448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,2048,0.00906577789121204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,2048,0.006776888751321369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,2048,0.013234666652149625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,1536,0.007847111258241866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,1536,0.0057911111248864075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,1536,0.010758221977286868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,1024,0.006805333412355847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,1024,0.004925333377387789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,1024,0.00927022182279163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,768,0.006246222390068903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,768,0.004451555510361989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,768,0.008700444466537898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,512,0.003940444439649582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,512,0.008022222254011366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,256,0.0033813334173626373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,256,0.007416889071464539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,512,0.005688888745175467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,256,0.005116444494989184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,128,0.003112888791494899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,128,0.007019555403126611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,64,0.0029395555870400537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,256,128,0.004839999808205498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,64,0.007151111132568783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,256,32,0.003031111011902491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,256,32,0.007145778172545963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,65536,0.04351111253102621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,65536,0.2316666709052192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,16384,0.01648177703221639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,16384,0.06046666701634725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,12288,0.013206221991115145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,65536,0.14263644483354357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,12288,0.04619910981920031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,10240,0.011928000383906893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,10240,0.04013333386845059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,16384,0.03943377733230591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,12288,0.030561778280470107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,8192,0.011337777806652917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,8192,0.033371554480658636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,8192,0.02209866709179348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,10240,0.026191110412279766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,7168,0.010811555716726514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,7168,0.029979556798934937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,6144,0.010109333528412713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,7168,0.020026667250527274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,6144,0.01789155602455139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,6144,0.026757333013746474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,5120,0.016100444727473788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,5120,0.009595555563767752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,5120,0.02392622166209751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,4096,0.008961777720186446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,4096,0.020097777247428894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,3584,0.009287999735938178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,3584,0.01848444508181678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,3072,0.008451555338170793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,3072,0.017062221964200337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,4096,0.013857777747843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,3584,0.012969777815871768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,3072,0.011146666275130378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,2560,0.010000000397364298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,2560,0.006792000184456508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,2560,0.015180443723996481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,2048,0.0058248887459437055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,2048,0.008791111409664154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,2048,0.01258933295806249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,1536,0.005063999858167437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,1536,0.007673777639865875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,1536,0.010623999767833285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,1024,0.006572444405820634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,1024,0.004332444320122401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,1024,0.009230222139093611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,768,0.003973333372010125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,768,0.00867555538813273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,512,0.005580444302823808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,512,0.003604444364706675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,512,0.00797155582242542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,256,0.0032426667296224167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,256,0.007357333269384172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,768,0.006051555689838197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,128,0.0047484445903036325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,128,0.002973333415057924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,128,0.007033777733643849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,64,0.0028159999185138275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,64,0.007072889142566257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,128,32,0.0029253332565228143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1024,128,256,0.004979555391603046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,128,32,0.007080888582600488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,65536,0.04487288991610209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,65536,0.23334222369723848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,16384,0.015378667248619927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,16384,0.06000622113545736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,12288,0.013089777694808112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,12288,0.0462080008453793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,10240,0.012157333393891653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,10240,0.040088001224729754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,8192,0.010624888870451184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,8192,0.033370667033725314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,7168,0.010091555615266165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,7168,0.029763556189007227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,6144,0.009545777406957414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,6144,0.02677333354949951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,5120,0.009015111459626092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,5120,0.02380888826317257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,4096,0.008450667063395182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,4096,0.020041777027977836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,3584,0.008336889247099558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,3584,0.0184542222155465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,3072,0.0075288886825243635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,3072,0.017039111918873258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,2560,0.006292444550328785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,2560,0.01499111122555203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,2048,0.005547555370463266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,2048,0.01237422227859497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,1536,0.005551110953092575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,1536,0.010637332995732626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,1024,0.004531555705600315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,1024,0.009290667043791877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,768,0.0041848888827694785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,768,0.00869688888390859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,512,0.0037537775933742523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,512,0.007924444145626491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,256,0.0033902221669753394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,256,0.00738844441043006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,128,0.0030542222989930045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,128,0.0069226668112807805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,64,0.0029528888149393927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,64,0.006992888947327931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,16384,0.06000889009899563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,64,32,0.003118222165438864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,10240,0.011623111036088733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,64,32,0.007053333024183909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,16384,0.014877332581414117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,65536,0.043177776866488986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,65536,0.2318720022837321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,12288,0.012684444586435953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,12288,0.04608177807595995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,10240,0.040004445446862116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,8192,0.010245333115259806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,8192,0.03347377644644843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,7168,0.009560888840092553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,7168,0.029792000850041706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,6144,0.008650666309727563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,6144,0.026733333865801494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,5120,0.009783999787436591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,5120,0.023732443650563557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,4096,0.008813333180215623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,4096,0.01993333299954732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,3584,0.007974222302436829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,3584,0.018481777773963082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,3072,0.0075235557225015424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,3072,0.016914667354689706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,2560,0.006173333360089197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,2560,0.015125332607163323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,2048,0.005511999958091312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,2048,0.01220355596807268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,1536,0.004837333328194088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,1536,0.010608889162540436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,1024,0.004118222329351637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,1024,0.009167110754383935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,128,0.002855111120475663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,768,0.003773333297835456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,768,0.008711111214425828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,512,0.003400000019205941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,512,0.007924444145626491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,256,0.003080888961752256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,256,0.007371555599901412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,128,0.0068693334857622785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,64,0.0026702222724755607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,64,0.007017778025733099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1024,32,32,0.0027191111197074256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1024,32,32,0.007033777733643849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,16384,1.2738444010416667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,16384,2.396043565538194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,16384,1.2243777381049263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,12288,0.9026293224758573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,12288,1.8234354654947917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,12288,0.8918551339043511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,10240,0.727000872294108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,10240,0.7382924291822645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,10240,1.5298115412394206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,6144,0.43747909863789874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,8192,0.5555911064147949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,8192,1.181836446126302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,7168,0.4931831359863281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,7168,0.9524106979370117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,8192,0.5958231290181478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,6144,0.8583466741773816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,7168,0.5257671144273546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,6144,0.41768889957004124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,5120,0.690482669406467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,4096,0.3028079933590359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,4096,0.28960710101657444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,5120,0.3565048906538222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,4096,0.5563039779663086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,5120,0.3683448897467719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,3584,0.4794658025105794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,3584,0.271142217848036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,3584,0.25583823521931964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,3072,0.4194186793433295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,3072,0.23048800892300078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,3072,0.22012000613742402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,2560,0.19786222775777182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,2560,0.34673065609402126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,2560,0.1854391098022461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,2048,0.1625866625044081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,2048,0.2840382258097331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,2048,0.15293778313530818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,1536,0.21718133820427787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,1536,0.12125689453548855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,1024,0.1499368879530165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,1536,0.12535288598802355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,1024,0.09321244557698567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,1024,0.08952889177534316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,768,0.11919289165072972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,768,0.07825599776373969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,512,0.08562311198976304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,768,0.07269155316882663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,512,0.06926133235295613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,512,0.055713779396480985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,256,0.04189599884880913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,256,0.05261244376500448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,256,0.05764000283347236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,65536,128,0.039919998910692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,128,0.042275554604000516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,128,0.05355644557211134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,64,0.037645333343082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,64,0.05537422166930305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,65536,32,0.03830577929814657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,65536,32,0.056052446365356445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,65536,1.1945315466986763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,65536,2.485896004570855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,65536,1.2540648778279622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,16384,0.3007066779666477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,16384,0.5445511076185439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,16384,0.2920622295803494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,12288,0.22741778691609701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,12288,0.3973848819732666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,12288,0.21971466806199816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,10240,0.18927733103434244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,10240,0.33325423134697807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,10240,0.182297772831387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,8192,0.2743244436052111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,8192,0.1479840013715956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,8192,0.15650222036573622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,7168,0.23856888877020943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,7168,0.1287742190890842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,7168,0.13360267215304905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,6144,0.11600622865888809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,6144,0.2053004503250122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,6144,0.11362577809227838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,5120,0.17500000529819063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,5120,0.09556800127029419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,4096,0.13724533716837564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,4096,0.07680977715386285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,5120,0.09991021951039632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,4096,0.0805635584725274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,3584,0.0717759993341234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,3584,0.12454222308264838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,3584,0.07004533211390178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,3072,0.10511022143893772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,3072,0.05941066477033827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,3072,0.06271466943952772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,2560,0.05417955583996243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,2560,0.09032622310850356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,2560,0.051791111628214516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,2048,0.04501510990990532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,2048,0.07510577970080905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,2048,0.04248977700869242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,1536,0.03606755534807841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,1536,0.05727377865049574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,1536,0.034456888834635414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,1024,0.04105777872933282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,1024,0.02753066685464647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,768,0.02328444520632426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,768,0.03331644336382548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,1024,0.027681777874628704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,768,0.024431111084090337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,512,0.02522666586769952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,512,0.021549334128697712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,512,0.019547555181715224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,256,0.014975999792416891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,256,0.01759733259677887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,256,0.01835555500454373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,128,0.0136835558546914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,16384,128,0.013179555535316467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,128,0.01590755581855774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,64,0.01148888882663515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,64,0.01609066625436147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,16384,32,0.011619555453459421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,16384,32,0.017280000779363845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,65536,0.8881155649820963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,16384,0.2292106681399875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,16384,0.39873954984876847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,16384,0.2357893255021837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,65536,0.9817991256713867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,65536,1.797919167412652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,12288,0.1738684442308214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,12288,0.30312267939249676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,12288,0.17689066463046602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,10240,0.14503555827670628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,10240,0.2508160008324517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,10240,0.15027733643849692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,8192,0.11767644352383083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,8192,0.1998506651984321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,8192,0.12234400378333198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,7168,0.17305421829223633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,7168,0.10545866356955634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,7168,0.1073911123805576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,6144,0.15225244892968073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,6144,0.0926284458902147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,5120,0.07490311066309611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,5120,0.1287626690334744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,6144,0.0892568892902798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,5120,0.07810844315422906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,4096,0.0629626644982232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,4096,0.10310133298238118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,4096,0.06416089004940458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,3584,0.05566755268308851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,3584,0.09195378091600205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,3584,0.05680888891220093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,3072,0.04860088891453213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,3072,0.07748711109161377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,3072,0.049885332584381104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,2560,0.04251377781232198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,2560,0.06672444608476427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,2560,0.043191999197006226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,2048,0.03547022077772353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,2048,0.0537395543522305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,2048,0.03617066807217068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,1536,0.04284977912902832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,1536,0.029076443778143987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,1536,0.029481778542200725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,1024,0.023060444328520033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,1024,0.03000266684426202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,1024,0.02373955481582218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,768,0.02398577829202016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,768,0.021163556310865615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,768,0.020219556159443326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,512,0.01850044396188524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,512,0.016952888833151925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,512,0.018805333309703402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,64,0.008387555678685507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,256,0.013870221873124441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,256,0.012897777888509961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,256,0.01567911108334859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,12288,128,0.011773332953453064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,128,0.009709333380063375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,128,0.013448889056841532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,64,0.014091556270917257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,12288,32,0.008757332960764566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,12288,32,0.015052444405025907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,65536,1.5609875784979925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,65536,0.8483297559950086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,65536,0.8219617737664117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,16384,0.3396560086144342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,16384,0.19159289201100668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,12288,0.15156977706485325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,12288,0.25914843877156574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,12288,0.1454791095521715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,10240,0.22128800551096597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,16384,0.19691644774542916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,10240,0.12846399678124323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,10240,0.12096888489193386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,8192,0.1753742165035672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,8192,0.09886311160193549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,8192,0.10327377584245469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,7168,0.09127377801471287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,7168,0.15648089514838323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,7168,0.08689510822296143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,6144,0.07961599694357978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,6144,0.13016088803609213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,6144,0.07558221949471368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,5120,0.06705244382222493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,5120,0.11152711179521348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,5120,0.06450844473308988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,4096,0.08893155389361912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,4096,0.053247110711203686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,4096,0.055530667304992676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,3584,0.07948444286982219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,3584,0.04984088738759359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,3584,0.047353777620527476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,3072,0.04365066687266032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,3072,0.0690551135275099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,3072,0.04139644569820828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,2560,0.05929599867926704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,2560,0.03625511129697164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,2048,0.047600001096725464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,2048,0.030280888080596924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,2560,0.03835377759403653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,1536,0.0365608897474077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,2048,0.03195377853181627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,1536,0.026313778426912095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,1536,0.024904888537194993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,1024,0.026338666677474976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,1024,0.020626667473051283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,768,0.018096889058748882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,768,0.021151110529899597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,768,0.018351111147138808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,512,0.016304888659053378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,1024,0.020976889464590285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,512,0.016283555163277518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,512,0.015274667077594332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,256,0.012229333321253458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,256,0.013202666408485837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,256,0.012883555557992725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,10240,128,0.0107004443804423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,128,0.011662222445011139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,128,0.01180622229973475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,64,0.010788444843557147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,64,0.012349333200189801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,10240,32,0.01108888867828581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,10240,32,0.013155555559529198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,65536,1.1051244735717773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,65536,0.6454479959275987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,16384,0.1580000056160821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,16384,0.27439289622836643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,65536,0.6111182106865777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,16384,0.15161689122517905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,12288,0.20941689279344347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,12288,0.11384088463253444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,10240,0.09909155633714463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,10240,0.17381510469648573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,10240,0.09580000241597493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,12288,0.1192373302247789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,8192,0.08291555775536431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,8192,0.13961511188083225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,8192,0.07891288730833265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,7168,0.0711839993794759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,7168,0.12077866660224067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,7168,0.06911200284957886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,6144,0.06243377923965454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,6144,0.10570755269792344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,6144,0.06117155816819933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,5120,0.053467555178536304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,5120,0.089610669347975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,5120,0.051353779104020864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,4096,0.04342044393221537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,4096,0.07220444414350721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,4096,0.042656001117494374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,3584,0.06472711430655585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,3584,0.038246221012539335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,3072,0.03525333272086249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,3072,0.053977777560551964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,3072,0.03364533185958862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,3584,0.03962044583426581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,2560,0.04626933402485318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,2560,0.030908445517222088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,2560,0.029375112719006006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,2048,0.026093333959579468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,2048,0.03825599948565165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,2048,0.025035555164019268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,1536,0.022272888157102797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,1536,0.02944799926545885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,1536,0.02104444470670488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,1024,0.021203555994563635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,1024,0.017335111896197002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,1024,0.01791911158296797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,768,0.01740977830357022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,768,0.015588444140222339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,768,0.016001777516471017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,512,0.013590222431553734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,512,0.013507555756303998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,256,0.011483555866612328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,256,0.00959022260374493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,256,0.01089244418674045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,128,0.010075555907355415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,8192,512,0.013705777625242868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,128,0.00737955586777793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,128,0.010092444717884064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,64,0.006404444575309753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,64,0.010608889162540436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,8192,32,0.006723555425802867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,8192,32,0.011032000184059143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,65536,0.9583297305636935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,65536,0.5982924567328559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,65536,0.5980266465081109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,16384,0.24513512187533906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,16384,0.14867644839816624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,16384,0.14607821570502386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,12288,0.1132791174782647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,12288,0.17851111623975965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,12288,0.11282132731543647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,10240,0.14896889527638754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,10240,0.09620710876252915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,10240,0.09430222378836738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,8192,0.07711111174689399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,8192,0.12244533167945014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,8192,0.07576888799667358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,7168,0.0684151119656033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,7168,0.10720622539520264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,7168,0.06678844160503812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,6144,0.09445155329174465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,6144,0.05861333343717787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,5120,0.07828355497784086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,6144,0.05992799997329712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,5120,0.04993244343333774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,4096,0.06231200032764011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,4096,0.04114933477507697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,5120,0.050944887929492526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,4096,0.0418542226155599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,3584,0.05573066737916735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,3584,0.037434667348861694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,3584,0.03847555650605096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,3072,0.033979554971059166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,3072,0.04728622237841288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,3072,0.03273422188229031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,2560,0.030180444320042927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,2560,0.04032444291644626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,2560,0.028667555914984807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,2048,0.02558311157756382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,2048,0.032855110036002264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,2048,0.024693333440356787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,1536,0.025926222403844196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,1536,0.02053600052992503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,1024,0.017443555924627516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,1024,0.018292443619834054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,1536,0.021893334057595994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,1024,0.0173342227935791
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,768,0.015050666199790107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,768,0.015459555718633862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,512,0.011676444775528379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,512,0.013086222112178802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,768,0.015676443775494892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,256,0.011230222052998014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,256,0.008685333033402761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,512,0.013541333377361298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,256,0.010858666565683154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,7168,128,0.009958222508430481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,128,0.007330666813585494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,128,0.010067555639478896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,64,0.006309333360857434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,64,0.010695999695195092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,7168,32,0.006587555425034628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,7168,32,0.010871999793582492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,65536,0.8157039748297797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,65536,0.6424160003662109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,65536,0.5744204521179199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,16384,0.19318932957119414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,16384,0.14574400583902994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,16384,0.15480977959103054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,12288,0.14790222379896376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,12288,0.10876444313261245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,12288,0.11728710598415798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,10240,0.12451910972595215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,10240,0.09297421905729507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,8192,0.07862044705284967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,8192,0.1006773312886556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,8192,0.0759048859278361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,10240,0.09951733218299018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,7168,0.07023821936713324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,7168,0.08909244669808282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,7168,0.06693955262502034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,6144,0.061025778452555336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,6144,0.07534222470389472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,6144,0.05795644389258491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,5120,0.05234666665395101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,5120,0.06511822011735705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,5120,0.049080888430277504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,4096,0.05235911077923245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,4096,0.04317422376738655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,4096,0.04099999864896139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,3584,0.046145776907602944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,3584,0.03686755564477708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,3584,0.03882222374280294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,3072,0.03997866643799676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,3072,0.032636443773905434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,3072,0.03461155626508925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,2560,0.03405244482888116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,2560,0.028560890091790095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,2048,0.02755111124780443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,2048,0.024247111545668706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,2560,0.0307262208726671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,1536,0.02221511138810052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,2048,0.02607288956642151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,1536,0.02052977846728431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,1536,0.02161866592036353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,1024,0.016509332590632968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,1024,0.017072000437312655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,768,0.013712888790501488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,1024,0.01792266633775499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,768,0.015818667080667283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,768,0.015260444747077094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,512,0.010910222099887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,512,0.012528888881206512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,256,0.01087288889620039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,256,0.007531555162535773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,512,0.013228444589508904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,256,0.010762666662534079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,128,0.006249777972698212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,128,0.009959999885823991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,64,0.00544177782204416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,6144,128,0.009438222481144799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,64,0.010568000376224518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,6144,32,0.0057653333577844836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,6144,32,0.010631110933091907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,65536,0.416362656487359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,65536,0.7110657691955566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,65536,0.4075484540727403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,16384,0.17507733239067924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,16384,0.10308977630403306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,12288,0.12859022617340088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,12288,0.07853955692715116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,10240,0.11070844199922349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,10240,0.06691110796398587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,16384,0.10674044158723618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,12288,0.0828382240401374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,10240,0.06977778010898165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,8192,0.0565777752134535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,8192,0.08705688847435845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,8192,0.054414222637812294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,7168,0.05057600140571594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,7168,0.07864533530341254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,7168,0.04800622330771553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,6144,0.06695110930336846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,6144,0.04426133301522997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,6144,0.042583111259672374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,5120,0.05684622128804525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,5120,0.03864266806178623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,5120,0.036686221758524575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,4096,0.04535911149448819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,4096,0.031001776456832886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,4096,0.032456000645955406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,3584,0.029430223835839167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,3584,0.04061955544683669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,3584,0.028027554353078205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,3072,0.02681866619322035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,3072,0.035517334938049316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,3072,0.024950222836600408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,2560,0.03050311075316535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,2560,0.023959999283154804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,2560,0.0222053329149882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,2048,0.020125332805845473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,2048,0.025049777494536504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,2048,0.01939377850956387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,1536,0.017356443736288283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,1536,0.01977688901954227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,1536,0.016504888733228047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,1024,0.015001777145597668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,1024,0.014520888527234396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,1024,0.013972444666756524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,768,0.013080889152155982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,768,0.01253600004646513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,768,0.012243555651770698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,512,0.011437333292431302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,512,0.010133333504199982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,512,0.010068444742096795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,256,0.009676444033781687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,256,0.007096889118353526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,256,0.009143110778596666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,128,0.0058906666106647914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,128,0.00847822262181176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,5120,128,0.00870488915178511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,64,0.005507555686765247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,64,0.008964444200197855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,5120,32,0.005659555395444234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,5120,32,0.009125333693292405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,65536,0.33345511224534774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,65536,0.5429609086778429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,16384,0.08588000138600667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,16384,0.13868267006344265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,16384,0.10080622302161323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,65536,0.38963998688591855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,12288,0.06590311394797431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,12288,0.10358222325642903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,12288,0.07648888561460707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,10240,0.055742224057515465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,10240,0.08532000250286526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,10240,0.06502755482991536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,8192,0.06938755512237549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,8192,0.05301600032382541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,7168,0.040970666540993586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,7168,0.06181244717703926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,7168,0.04703288939264086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,8192,0.04600711001290215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,6144,0.053708444039026894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,6144,0.036080890231662326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,6144,0.04112799962361654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,5120,0.044799112611346774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,5120,0.03156177865134345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,5120,0.03572888837920295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,4096,0.03696977761056688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,4096,0.030271109607484605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,3584,0.024474667178259954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,3584,0.03188000122706095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,3584,0.027474666635195415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,4096,0.026464000344276428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,3072,0.028098666005664404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,3072,0.022031999296612207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,2048,0.019017777509159513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,3072,0.024667556087176006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,2560,0.023987554841571387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,2560,0.019778667224778067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,2560,0.0221413340833452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,2048,0.016823111308945548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,2048,0.019902222686343722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,1536,0.014800889624489678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,1536,0.015997333659066092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,1536,0.01625866691271464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,1024,0.012191111014948951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,1024,0.012443555725945367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,1024,0.013674666484196981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,768,0.01145866678820716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,768,0.010325333310498131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,768,0.011733333269755045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,512,0.010326222413116032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,512,0.008020444048775567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,512,0.010001777774757808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,256,0.00888888869020674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,256,0.005997333261701796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,256,0.00907022257645925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,128,0.008447111480765873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,128,0.005070222334729301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,64,0.004512888689835866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,64,0.008823999928103553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,4096,32,0.0046951112647851305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,4096,32,0.009036444127559662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,4096,128,0.007838221887747446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,65536,0.31603821118672687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,65536,0.4696897930569119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,65536,0.3895768854353163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,16384,0.08247733116149902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,16384,0.12150311470031738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,16384,0.10019199715720283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,12288,0.06266044245825873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,12288,0.09119555685255264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,12288,0.07613244321611193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,10240,0.05389066537221273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,10240,0.07789333661397298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,10240,0.06405333015653822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,8192,0.060681780179341636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,8192,0.0524453322092692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,8192,0.0442808899614546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,7168,0.05443911088837517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,7168,0.0467564430501726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,6144,0.047115554412206016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,6144,0.04139288928773668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,5120,0.030335999197430078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,4096,0.0325271122985416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,5120,0.04014133413632711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,7168,0.039601776334974505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,5120,0.03587199913130866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,6144,0.03471555643611484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,4096,0.025926222403844196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,4096,0.030258665482203167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,3584,0.028990222348107234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,3584,0.02747022277779049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,3072,0.02141777839925554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,3584,0.023581332630581323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,3072,0.02533955540921953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,3072,0.02460711035463545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,2560,0.019359111785888672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,2560,0.021723555194007024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,2560,0.021870222356584337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,2048,0.018256889449225534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,2048,0.019162666466501024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,1536,0.01498222185505761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,1536,0.01626044511795044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,2048,0.016604445046848722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,1024,0.012154666913880242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,1536,0.014498665928840637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,1024,0.011666666302416058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,1024,0.013032000097963544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,768,0.011266666981909009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,768,0.010038221875826517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,768,0.01124533348613315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,512,0.010136889086829292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,512,0.008126222425036961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,512,0.010031110710567897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,256,0.008351999852392409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,256,0.0062773335311147906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,256,0.009148444566461775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3584,128,0.007847111258241866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,128,0.005447111195988125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,128,0.008464888566070134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,64,0.00424177779091729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,64,0.008893333375453949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3584,32,0.004327110946178436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3584,32,0.008941333327028487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,65536,0.322613345252143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,65536,0.4101395606994629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,65536,0.38745954301622176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,16384,0.10236977868609959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,16384,0.09902489185333252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,12288,0.07902489105860393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,16384,0.08495999707116021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,12288,0.06384978029463026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,12288,0.076273779074351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,10240,0.06574400266011556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,10240,0.05435466766357422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,10240,0.0644435551431444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,8192,0.044666667779286705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,8192,0.05349777804480659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,8192,0.05258311165703667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,7168,0.04636355572276645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,7168,0.04709600077735054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,7168,0.040088001224729754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,6144,0.03526399864090814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,6144,0.03994044330384996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,6144,0.040957334968778826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,5120,0.03464355402522617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,5120,0.03560622202025519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,4096,0.02602311140961117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,4096,0.028156446086035833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,4096,0.02990488873587714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,3584,0.02374222212367588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,3584,0.025007999605602686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,3584,0.02734666731622484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,3072,0.021528888079855178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,3072,0.02167644434505039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,3072,0.024400888217820063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,5120,0.030905776553683813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,2560,0.018597332967652213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,2560,0.02165866725974613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,2048,0.015565334094895257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,2048,0.018984888990720112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,1536,0.014389332797792224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,1536,0.012857777376969656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,1536,0.01622488929165734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,1024,0.010001777774757808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,1024,0.012503110700183444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,768,0.011360000405046674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,768,0.008579555484983657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,768,0.010991111397743225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,512,0.009916444619496664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,2560,0.019433778193261888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,512,0.006673777682913675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,2048,0.01682577696111467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,1024,0.01256711118751102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,512,0.009964444571071202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,256,0.007956444389290279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,256,0.005234666582610872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,256,0.009036444127559662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,128,0.004550222307443619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,128,0.008300444318188561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,64,0.004044444610675176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,64,0.008683555656009251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,3072,128,0.007147555549939473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,3072,32,0.004263999975389904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,3072,32,0.008769777913888296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,65536,0.35093243916829425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,65536,0.23388621542188856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,16384,0.08995378017425537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,16384,0.06013511286841499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,65536,0.2465368906656901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,16384,0.06393688917160034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,12288,0.049810667832692467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,12288,0.06807200113932292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,12288,0.04683199856016371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,10240,0.04257066713439094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,10240,0.05730311075846354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,10240,0.0399617784553104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,8192,0.045575999551349215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,8192,0.03350044290224711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,7168,0.031586666901906334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,7168,0.04087377919091119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,7168,0.03024266825781928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,6144,0.03586488962173462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,6144,0.026995556222067937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,5120,0.02482044365670946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,5120,0.030015998416476782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,5120,0.02405066622628106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,4096,0.021008888880411785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,4096,0.02474844455718994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,4096,0.020610666937298246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,3584,0.01942044496536255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,8192,0.03526399864090814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,6144,0.027952889601389568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,3584,0.021948445174429152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,3584,0.019167999426523846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,3072,0.019341333044899833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,3072,0.01758044461409251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,2560,0.016699555847379897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,2560,0.015944888194402058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,3072,0.01773066653145684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,2048,0.01423822177780999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,2048,0.01423288881778717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,2560,0.016172443826993305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,1536,0.011711999773979187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,1536,0.012741333080662621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,1024,0.010626667075686984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,2048,0.013836444252067141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,1536,0.012328888807031842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,1024,0.00926755534278022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,1024,0.010151999692122141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,768,0.007789333661397298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,768,0.008999110923873054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,512,0.006170666466156642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,512,0.008288000192907121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,256,0.004956444518433677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,256,0.007629333270920648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,768,0.009257777697510189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,128,0.006065777606434292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,512,0.007910221815109253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2560,256,0.006897777732875612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,128,0.004280000097221798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,128,0.007163555257850223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,64,0.003954666770166821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,64,0.0073902226156658595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2560,32,0.004054222255945206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2560,32,0.0075297777851422625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,65536,0.2749679883321126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,65536,0.22137778335147432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,16384,0.06316622098286946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,16384,0.07784710990058051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,16384,0.05786844756868151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,65536,0.23923110961914062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,12288,0.055919110774993896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,12288,0.04545333319240146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,12288,0.04863199922773573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,10240,0.041478223270840116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,10240,0.048756443791919284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,10240,0.03861777649985419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,8192,0.03465422325664096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,8192,0.04194577866130405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,8192,0.03240266773435805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,7168,0.031521777311960854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,7168,0.03479555580351088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,7168,0.02929511004024082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,6144,0.030282666285832722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,6144,0.026519111461109583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,5120,0.025519111090236243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,6144,0.027777777777777776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,5120,0.023746665981080797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,4096,0.02273599969016181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,5120,0.024600888291994732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,4096,0.020336000455750358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,3584,0.019556444552209642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,3584,0.018734221657117207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,3072,0.017068444026841056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,3072,0.017299556069903903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,4096,0.02070755594306522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,3584,0.0192239996459749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,3072,0.01772799922360314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,2560,0.01612355477280087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,2560,0.013595555391576556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,2560,0.015645333462291293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,2048,0.011692444483439127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,2048,0.014107555150985718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,1536,0.009754666851626502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,1536,0.012235555383894177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,1024,0.010035555395815108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,1024,0.00755555596616533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,1024,0.009786666267448002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,768,0.006588444527652528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,768,0.008864888714419471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,2048,0.01366666704416275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,1536,0.012082666986518435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,512,0.007378666765160031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,768,0.0087013335691558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,512,0.005330666485759947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,512,0.008221333225568136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,256,0.004461333155632019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,256,0.006287999865081575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,256,0.0075706665714581805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,128,0.003957333250178231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,128,0.0070426662762959795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,64,0.003560888684458203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,2048,128,0.005712000032265981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,64,0.007274666594134436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,2048,32,0.003765333443880081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,2048,32,0.00741511086622874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,65536,0.20805688699086508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,16384,0.06282133526272245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,16384,0.06038844585418701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,65536,0.21265955766042074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,16384,0.05743910868962606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,12288,0.0473599996831682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,12288,0.04490310947100321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,10240,0.04152977797720168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,10240,0.03807555635770162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,65536,0.2317377726236979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,8192,0.03500888744990031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,12288,0.049004445473353066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,10240,0.04187733266088697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,8192,0.030604445272021826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,8192,0.0320355561044481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,7168,0.026644445127911035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,7168,0.029014223151736792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,7168,0.03197244471973843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,6144,0.027658666173617046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,6144,0.023416888382699754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,6144,0.02624977793958452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,5120,0.024361777636739943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,5120,0.020071110791630216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,5120,0.023249778482649062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,4096,0.02051555613676707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,4096,0.016823111308945548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,4096,0.020075556304719713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,3584,0.015057777365048727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,3584,0.01865688959757487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,3072,0.017314665847354464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,3072,0.013398222625255585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,3584,0.019089778264363606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,3072,0.01683111157682207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,2560,0.012988444831636218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,2560,0.015786666009161208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,2560,0.015465777781274585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,2048,0.01127466642194324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,2048,0.013586666848924426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,2048,0.013929777675204806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,1536,0.00962311112218433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,1536,0.01186044431395001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,1536,0.011549333731333414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,1024,0.0073902226156658595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,1024,0.009415999882751042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,768,0.006214222146405115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,768,0.008760000268618265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,512,0.0053004444473319584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,512,0.008080888953473832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,1024,0.009879999690585667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,256,0.004259555704063839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,768,0.00850577818022834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,512,0.007106666763623555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,256,0.0074755557709270054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,256,0.006039111150635614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,128,0.0038648889296584656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,128,0.007073777417341868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,64,0.003463111196955045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,64,0.007184888753626082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1536,32,0.003544888976547453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1536,32,0.007204444044166141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,65536,0.1863084501690335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,65536,0.14860088295406765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,65536,0.21262755658891466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1536,128,0.005620444400442972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,16384,0.04144088758362664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,16384,0.05651466714011299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,16384,0.05176800158288744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,12288,0.0328453348742591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,12288,0.04381688767009311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,10240,0.03340444299909804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,10240,0.028352889749738906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,10240,0.03792533278465271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,12288,0.03938755393028259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,8192,0.03143999973932902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,8192,0.03162577748298645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,7168,0.025231111380789015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,7168,0.026778666509522334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,7168,0.029150221082899306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,6144,0.023457777169015672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,6144,0.026216889421145122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,5120,0.01981599960062239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,5120,0.02013688948419359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,5120,0.023006222314304773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,8192,0.02774577836195628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,4096,0.018235555953449674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,6144,0.022460444105996027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,4096,0.019887111253208585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,4096,0.016892444756295946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,3584,0.015722667177518208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,3584,0.015196444259749519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,3584,0.018370666437678866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,3072,0.014343111051453484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,3072,0.013234666652149625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,3072,0.016881777180565726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,2560,0.013148444394270578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,2560,0.009522666533788046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,2560,0.015346666177113851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,2048,0.007847111258241866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,2048,0.010943111446168689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,2048,0.013218666944238873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,1536,0.009330666727489894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,1536,0.00674044465025266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,1536,0.01092088884777493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,1024,0.007511999871995714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,1024,0.0053288886944452924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,1024,0.009190222455395592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,768,0.00490488898422983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,768,0.006787555499209299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,768,0.008590222232871586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,512,0.004384888956944148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,512,0.00796888851457172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,256,0.0038204445607132385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,256,0.007368000017272101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,128,0.005137777576843898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,512,0.006247111078765657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,1024,256,0.005565333283609814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,128,0.003479111111826367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,128,0.007039111521508958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,64,0.0032222221295038858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,64,0.007101332975758447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,1024,32,0.0032835555159383347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,1024,32,0.007167110840479533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,65536,0.11117333836025661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,65536,0.21173422866397432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,16384,0.04624266756905449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,16384,0.03370044297642178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,16384,0.05650577942530314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,12288,0.03542755709754096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,12288,0.027436445156733196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,12288,0.043302221430672534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,10240,0.03084889054298401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,65536,0.1654640038808187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,10240,0.024546666277779475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,10240,0.03763555486996969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,8192,0.025633777181307476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,8192,0.019031110737058852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,8192,0.031403554810418025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,7168,0.017296888762050204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,7168,0.028780443800820246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,6144,0.015336889359686108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,6144,0.025722665919197932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,7168,0.02365422248840332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,5120,0.013547555440002017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,5120,0.02276355524857839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,6144,0.02066044509410858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,5120,0.01830755505296919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,4096,0.011544889046086205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,4096,0.019658666517999437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,4096,0.015607110328144498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,3584,0.010671111444632212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,3584,0.01810933318403032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,3072,0.009539555344316695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,3072,0.016698666744761996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,2560,0.010520888699425591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,3584,0.014631999863518609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,3072,0.01333688861793942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,2560,0.015205333630243937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,2048,0.009226666556464301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,2048,0.012889777620633444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,2560,0.012229333321253458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,1536,0.008803555534945594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,1536,0.007435555259386699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,1536,0.010593777729405297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,1024,0.007611555357774098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,2048,0.010267555713653564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,1024,0.006005333529578314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,1024,0.009186666872766282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,768,0.004661333229806689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,768,0.0069191112286514705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,768,0.008650666309727563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,512,0.004044444610675176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,512,0.006112888869312074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,512,0.007958222594526079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,256,0.0035786665976047516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,256,0.007367110914654202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,256,0.005422222117582957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,128,0.003293333368168937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,128,0.006978666616810693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,64,0.0030462222380770575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,65536,0.07936977677875094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,64,0.007072000039948358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,768,32,0.003116444374124209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,768,32,0.007081777685218387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,768,128,0.005104889058404499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,65536,0.15686755710177952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,16384,0.04303822252485487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,16384,0.025506666964954797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,65536,0.2114915582868788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,16384,0.05598400036493937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,12288,0.03345155715942383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,12288,0.02072533302836948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,12288,0.04336888922585381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,10240,0.019052444232834708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,10240,0.037448889679378934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,8192,0.01663644446267022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,8192,0.0315324432320065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,7168,0.021987555755509272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,10240,0.028577778074476454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,8192,0.024021334118313257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,7168,0.015238222148683337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,7168,0.02864711152182685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,6144,0.01942933268017239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,6144,0.013471999930010902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,6144,0.02555199960867564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,5120,0.017465778523021273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,5120,0.011904000408119626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,5120,0.022767111659049988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,4096,0.014833778142929077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,4096,0.010436444646782346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,4096,0.01941688855489095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,3584,0.009677333136399588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,3584,0.018060444129837882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,3072,0.008756444685988957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,3072,0.016516443755891588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,2560,0.010928000013033548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,2560,0.00793866647614373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,3584,0.013659555878904132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,3072,0.01260444439119763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,2560,0.014940443966123792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,2048,0.006867555694447623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,2048,0.012714666624863943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,1536,0.0057591112951437635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,2048,0.009543110926946005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,1536,0.010464000205198923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,1536,0.008177777959240807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,1024,0.004781333522664176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,1024,0.007158222297827403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,1024,0.00907733374171787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,768,0.004410666724046071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,768,0.008525333470768398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,512,0.005942222144868638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,512,0.003910222401221593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,768,0.006507555643717448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,512,0.007858666280905405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,256,0.0034551111360390982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,256,0.007287110719415877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,256,0.005330666485759947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,512,128,0.005000888887378904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,128,0.0032088889016045462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,128,0.006919999917348226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,64,0.003039999968475766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,16384,0.03559022148450216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,64,0.00699733363257514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,512,32,0.0030871110243929755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,512,32,0.007051555646790399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,65536,0.12730488512251112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,65536,0.04605688982539707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,65536,0.21057155397203234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,16384,0.017478222648302715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,16384,0.05571644173728096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,12288,0.027986665566762287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,12288,0.014530667000346713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,12288,0.043284446001052856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,10240,0.014762666490342883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,10240,0.037802666425704956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,8192,0.013167999684810638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,8192,0.03141244583659702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,10240,0.02407288882467482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,7168,0.012392000191741519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,8192,0.02049955560101403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,7168,0.02847111225128174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,6144,0.011691555380821228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,6144,0.02556444373395708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,5120,0.010695111420419482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,5120,0.02265600032276577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,7168,0.01870133313867781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,4096,0.009943111075295342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,4096,0.019331556227472093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,6144,0.01661600006951226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,5120,0.01497155593501197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,3584,0.008908444808589088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,4096,0.012818666795889536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,3584,0.017847999930381775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,3584,0.01165422217713462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,3072,0.008072888685597314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,3072,0.016388444436921012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,2560,0.0069857777820693115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,3072,0.010565333068370819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,2560,0.014069333672523499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,2048,0.005853333406978183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,2560,0.009570666485362584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,2048,0.011907555162906647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,2048,0.00832177781396442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,1536,0.005123555660247803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,1536,0.010432889064153036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,1024,0.004381333374314838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,1024,0.00907022257645925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,512,0.0036888888312710654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,1536,0.00740444411834081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,1024,0.006401777681377199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,768,0.004051555362012652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,768,0.00853066643079122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,768,0.00601244428091579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,512,0.007810666329330868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,256,0.0032799999333090256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,256,0.007214222517278459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,512,0.005554666535721884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,128,0.002971555623743269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,256,0.005007110950019625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,128,0.006883555402358373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,64,0.0028355556229750314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,64,0.0069413334131240845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,256,32,0.002895999906791581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,256,32,0.00700800038046307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,65536,0.12519644366370306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,65536,0.03733333283000522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,16384,0.03469689024819268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,65536,0.21360977490743002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,16384,0.014283556077215405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,16384,0.05570311016506619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,12288,0.027432888746261597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,256,128,0.004779555731349521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,12288,0.012040889097584618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,12288,0.043770667579438954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,10240,0.011027555498811932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,10240,0.037808001041412354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,8192,0.009923555784755284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,8192,0.031368888086742826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,10240,0.023559111687872145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,7168,0.009642666412724389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,8192,0.01998666591114468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,7168,0.028588443994522095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,7168,0.018328888548745047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,6144,0.016302223006884258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,6144,0.008986666798591614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,6144,0.0255031105544832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,5120,0.009993777506881291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,5120,0.02270222206910451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,4096,0.008991999758614434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,5120,0.014480888843536377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,4096,0.019327110714382596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,3584,0.008120888637171851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,3584,0.01788444485929277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,3072,0.007696888513035244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,3072,0.015537778536478678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,2560,0.009101333717505137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,4096,0.012255999777052136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,2560,0.006355555521117316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,2560,0.013933334085676404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,2048,0.008176888856622908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,2048,0.005657777604129579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,3584,0.011400889191362592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,3072,0.009954666925801171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,2048,0.011704000333944956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,1536,0.004967111266321606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,1536,0.010334221853150262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,1024,0.004259555704063839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,1024,0.009026666482289633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,768,0.003909333298603694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,768,0.008464888566070134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,1536,0.007102222078376346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,1024,0.006190222170617845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,512,0.003539555602603488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,512,0.00775111135509279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,256,0.004921777794758479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,768,0.005869333528810077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,512,0.005370666583379109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,256,0.0031635556370019913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,256,0.007173333730962541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,768,128,128,0.0046826667255825466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,128,0.0029031110720502008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,128,0.006851555572615729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,64,0.0027760000278552375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,64,0.006888888776302338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,128,32,0.0028728888266616394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,128,32,0.0068853331936730286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,65536,0.036737776464886136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,16384,0.01276533305644989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,65536,0.21387822098202178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,16384,0.05590755409664578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,12288,0.011045333411958484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,12288,0.04379822148217095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,10240,0.010231110784742568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,10240,0.03751822312672933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,8192,0.00925244473748737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,8192,0.03126666612095303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,7168,0.008936000367005667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,7168,0.028544889556037054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,6144,0.008466666771305932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,6144,0.02549955579969618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,5120,0.008572444319725037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,5120,0.022555554906527202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,4096,0.007374222079912822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,4096,0.019202666150199044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,3584,0.00758844448460473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,3584,0.017792888813548617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,3072,0.007241778075695038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,3072,0.015587555037604438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,2560,0.006280888699822956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,2560,0.013401777380042605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,2048,0.005567111074924469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,2048,0.011719110939237805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,1536,0.004908444566859139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,1536,0.010310221877362993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,1024,0.0041946665280395085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,1024,0.008994667066468133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,768,0.0038684445122877755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,768,0.008435555630260045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,512,0.003469333259595765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,512,0.007751999629868402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,256,0.0031519999934567344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,256,0.007096889118353526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,128,0.00286311118139161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,128,0.006758222149478064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,64,0.002756444530354606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,64,0.00683822234471639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,64,32,0.0027786667148272195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,64,32,0.0068764446510208984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,65536,0.03443377878930833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,16384,0.012343111137549082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,65536,0.2140106625027127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,16384,0.055626663896772593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,12288,0.010647111468844943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,12288,0.043474667602115206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,10240,0.01000088867213991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,10240,0.03756711218092177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,8192,0.00926755534278022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,8192,0.031211555004119873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,7168,0.008669333325492011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,7168,0.028509332074059382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,6144,0.00907199995385276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,6144,0.02537600033813053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,5120,0.008417777717113495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,5120,0.0225600004196167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,4096,0.007183111376232571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,4096,0.01924088928434584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,3584,0.007238222493065729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,3584,0.01776444415251414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,3072,0.006873777757088344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,3072,0.015259555644459195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,2560,0.006164444403515921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,1024,0.008978666530715095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,2560,0.013219555219014486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,2048,0.005481777919663324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,512,0.0033760000434186724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,2048,0.011681777735551199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,1536,0.0047973332305749255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,1536,0.010325333310498131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,1024,0.004101333518822988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,768,0.0037413334680928122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,768,0.008419555922349295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,512,0.00777599960565567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,256,0.0030497778207063675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,256,0.007133333219422235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,128,0.0028311111446883944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,128,0.00674933319290479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,64,0.0026862221873468826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,64,0.006807111203670502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,768,32,32,0.002716444432735443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,768,32,32,0.006797333144479328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,16384,0.7552977667914497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,16384,1.68451107872857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,12288,1.1810515721638997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,16384,0.8382897906833224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,12288,0.5839342011345757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,10240,0.49425421820746523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,12288,0.5652933120727539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,10240,1.012812402513292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,10240,0.46958133909437394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,8192,0.7432444360521105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,8192,0.39230932129753965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,8192,0.3978417714436849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,7168,0.34668710496690536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,7168,0.6712453100416395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,7168,0.3268062273661296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,6144,0.3057360119289822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,6144,0.5501555336846246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,6144,0.2892471154530843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,5120,0.25792000028822154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,5120,0.47766579522026914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,4096,0.20693422688378227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,5120,0.2386559910244412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,4096,0.37089067035251194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,4096,0.19680444399515787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,3584,0.3314195473988851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,3584,0.17501866817474365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,3072,0.28711644808451336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,3072,0.14744533432854548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,2560,0.24337421523200142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,2560,0.12589866585201687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,3584,0.18001511361863878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,3072,0.1575288905037774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,2560,0.13279822137620714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,2048,0.11273954974280463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,2048,0.1964684459898207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,2048,0.10245511266920303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,1536,0.14997600184546575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,1536,0.08810133404201931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,1536,0.08156977759467231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,768,0.05352177884843615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,1024,0.06392621994018555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,1024,0.10323911243014866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,1024,0.06415733363893297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,768,0.05100711186726888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,768,0.0822168919775221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,512,0.04042488998836941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,512,0.059545775254567467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,512,0.04708088768853081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,256,0.031353778309292264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,256,0.03805333375930786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,256,0.039911111195882164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,128,0.03091111116939121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,128,0.03752977649370829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,64,0.02603911028967963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,64,0.03943822118971083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,65536,32,0.028945777151319716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,65536,32,0.040750222073660955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,65536,0.8444319831000434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,65536,128,0.028491556644439697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,65536,1.559159066942003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,65536,0.8148710992601184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,16384,0.20070489247639975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,16384,0.36557687653435605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,16384,0.1952435572942098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,12288,0.28141244252522785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,12288,0.14854844411214194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,10240,0.13162933455573186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,10240,0.23242134518093535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,12288,0.15149777465396458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,10240,0.12212977144453262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,8192,0.10537422365612453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,8192,0.1853875584072537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,8192,0.10235555304421319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,7168,0.16180888811747232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,7168,0.08933866686291164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,6144,0.1378435558742947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,6144,0.07672444317075941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,5120,0.11955732769436306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,5120,0.06438666582107544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,7168,0.0934257772233751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,6144,0.07923555374145508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,5120,0.06849777698516846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,4096,0.09314844343397354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,4096,0.05307911170853508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,4096,0.05691466728846232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,3584,0.051143109798431396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,3584,0.08319377899169922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,3584,0.04718666606479221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,3072,0.04393066631423103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,3072,0.07212089167700873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,3072,0.041119111908806696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,2560,0.03848888807826572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,2560,0.06182577874925402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,2560,0.036074668169021606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,2048,0.04999822378158569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,2048,0.030193779203626845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,1536,0.026833777626355488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,1536,0.0384897788365682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,2048,0.03276088833808899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,1536,0.024640000528759424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,1024,0.027899556689792212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,1024,0.0199946661790212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,1024,0.021448888712459143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,768,0.022055110997623865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,768,0.018773333893881906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,768,0.018179555733998615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,512,0.016660445266299777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,512,0.01587288909488254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,256,0.012669333153300814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,256,0.01180355581972334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,512,0.015896888242827523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,256,0.013592888911565145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,16384,128,0.011009777585665384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,128,0.008991999758614434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,128,0.011601777540312873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,64,0.007311999797821045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,64,0.012153777811262341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,16384,32,0.007704888780911763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,16384,32,0.013002666334311167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,65536,1.2143804762098525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,65536,0.601845317416721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,65536,0.6229795349968804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,16384,0.28727732764350045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,16384,0.15001332759857178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,12288,0.21372444099850127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,12288,0.11242666509416367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,10240,0.09772533178329468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,10240,0.18099110656314424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,10240,0.09604533513387044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,16384,0.15857422351837158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,12288,0.11752533912658691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,8192,0.14354044861263698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,8192,0.07599111398061116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,7168,0.07025955782996283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,8192,0.08119822210735746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,7168,0.12744177712334528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,7168,0.06907911433113946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,6144,0.06290577517615424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,6144,0.10914044247733222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,6144,0.0591191119617886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,5120,0.05191377798716227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,5120,0.09047466516494751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,5120,0.05093510945638021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,4096,0.07416711250940959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,4096,0.042040887806150645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,4096,0.04356177647908529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,3584,0.03939911060863071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,3584,0.06481866704093085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,3584,0.03726044297218323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,3072,0.057024889522128634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,3072,0.033089776833852134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,3072,0.03388888968361749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,2560,0.04780711068047417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,2560,0.02887111239963108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,2560,0.030116445488399927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,2048,0.025988444685935974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,2048,0.038607110579808555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,2048,0.024522667129834492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,1536,0.029868443806966145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,1536,0.022051556242836848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,1536,0.020468443632125854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,1024,0.018082666728231642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,1024,0.021686222818162706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,512,0.013588444226317935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,1024,0.016992888516849942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,768,0.016006223029560514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,768,0.017449777987268236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,768,0.01535288823975457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,512,0.013907555076811047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,512,0.013438222308953604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,256,0.009553777674833933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,256,0.010799110763602786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,256,0.011589333415031433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,12288,128,0.010101333260536194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,128,0.007287110719415877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,128,0.009861333502663506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,64,0.006130666782458623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,64,0.010586666564146677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,12288,32,0.006487111250559489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,12288,32,0.011108444796668159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,65536,0.9862533145480685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,65536,0.5625839763217503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,65536,0.5349964565700954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,16384,0.1327928834491306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,16384,0.23139556248982748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,16384,0.1400337749057346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,12288,0.10214844677183364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,12288,0.17275378439161512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,12288,0.1083128915892707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,10240,0.08809510866800944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,10240,0.14789066049787733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,10240,0.09095911184946696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,8192,0.0705804427464803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,8192,0.11728799343109131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,8192,0.07451999849743314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,7168,0.10102399852540757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,7168,0.06554666492674086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,6144,0.08735999796125625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,6144,0.05677511294682821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,7168,0.06172977553473579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,5120,0.07490222321616279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,6144,0.05483822358979119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,5120,0.04859288864665561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,4096,0.059697780344221324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,5120,0.046089778343836464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,4096,0.0388080014122857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,4096,0.040200889110565186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,3584,0.05340800020429823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,3584,0.03612000081274245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,3072,0.030453334252039593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,3584,0.03494577937655979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,3072,0.045533332559797496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,3072,0.03192977772818671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,2560,0.027080888549486797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,2560,0.03931022352642483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,2560,0.028121779362360638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,2048,0.03176177872551812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,2048,0.023930667175187006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,2048,0.023370666636361018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,1536,0.025116443634033203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,1536,0.019846222466892667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,1536,0.02010311186313629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,1024,0.016075554821226332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,1024,0.018430221411916945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,1024,0.016833777228991192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,768,0.01519466605451372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,768,0.01510933372709486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,512,0.01202222208182017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,512,0.013295110729005603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,256,0.010601777997281818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,256,0.008739555875460306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,256,0.010585777461528778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,128,0.00927022182279163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,128,0.006802666518423293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,768,0.014483556151390076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,128,0.00980799976322386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,10240,512,0.01257511145538754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,64,0.0058142224119769196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,64,0.010462221999963125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,10240,32,0.006167111297448476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,10240,32,0.010796444283591377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,65536,0.41868265469868976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,16384,0.10680266883638169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,16384,0.18203911516401503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,16384,0.1005013320181105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,65536,0.7165724436442057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,65536,0.406094233194987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,12288,0.08356533447901408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,12288,0.13767733838823107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,12288,0.07909511195288764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,10240,0.06870222091674805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,10240,0.11573955747816299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,10240,0.06512088908089532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,8192,0.0936044454574585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,8192,0.05466577741834852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,7168,0.08319644133249919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,7168,0.0475466681851281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,6144,0.04400444361898634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,8192,0.056695110268063016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,6144,0.06950577762391832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,6144,0.04222311245070564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,5120,0.03808355662557814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,5120,0.06011199951171875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,5120,0.03575377662976583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,4096,0.0321048895517985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,4096,0.04801955487993029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,4096,0.030378666188981798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,7168,0.05041155550214979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,3584,0.029273778200149536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,3584,0.04327644573317634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,3584,0.027493332823117573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,3072,0.03711377912097507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,3072,0.02480977773666382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,2560,0.03141244583659702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,2560,0.021783111823929682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,2048,0.020319110817379422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,2048,0.026075555218590632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,3072,0.025829333398077223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,2560,0.023407111565272015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,2048,0.01899377836121453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,1536,0.020277332928445604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,1536,0.016152888536453247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,1024,0.015308444698651632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,1024,0.013619555367363824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,1536,0.017416000366210938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,1024,0.014358222484588623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,768,0.013248888982666863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,768,0.012814222110642327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,768,0.01240355521440506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,512,0.0101724440852801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,512,0.010410666465759277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,256,0.009941333697901832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,512,0.011709333293967776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,256,0.007363555332024892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,256,0.008776889079146916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,8192,128,0.008786666724416945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,128,0.005807999935415056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,128,0.008283555507659912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,64,0.00519822206762102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,64,0.008857777549160851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,8192,32,0.005411555369695027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,8192,32,0.009114666945404476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,65536,0.4100320074293349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,16384,0.10851199759377374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,16384,0.1663448943032159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,65536,0.6950462129380969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,65536,0.39281243748135036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,16384,0.0982453359497918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,12288,0.12931288613213435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,12288,0.07422755824195014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,10240,0.06993777884377374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,10240,0.1054533322652181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,10240,0.06466133064693876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,12288,0.08164088593588935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,8192,0.08639111121495564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,8192,0.05232622226079305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,8192,0.056602663464016385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,7168,0.05089155501789517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,7168,0.07665244738260905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,7168,0.045863999260796436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,6144,0.06526755624347262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,6144,0.040391110711627536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,5120,0.055452442831463285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,5120,0.03522933191723294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,4096,0.04440088735686409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,4096,0.02964799933963352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,6144,0.04424355427424113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,5120,0.038372443781958684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,4096,0.03272799981964959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,3584,0.03994399971432156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,3584,0.030059556166330974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,3584,0.02696977721320258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,3072,0.03467288944456313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,3072,0.02422044508987003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,3072,0.026199110680156287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,2560,0.02972444560792711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,2560,0.021515554851955835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,2048,0.024667556087176006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,2560,0.023680888944202002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,2048,0.018763555420769584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,2048,0.02038755516211192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,1536,0.01963733302222358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,1536,0.01607288916905721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,1536,0.017646221650971305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,1024,0.014997333288192749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,1024,0.014956444501876831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,1024,0.013520888984203339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,768,0.013527111046844058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,768,0.01254755589697096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,768,0.011805333197116852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,512,0.010120000276300643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,512,0.010099555883142684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,512,0.011787555283970304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,256,0.006986666884687211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,256,0.010009778042634329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,256,0.008854221966531541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,128,0.005695999910434087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,128,0.008200888832410177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,64,0.004895110925038655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,64,0.008819555242856344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,7168,32,0.00544799988468488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,7168,32,0.009006222089131674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,7168,128,0.008962666822804345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,16384,0.08150400055779351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,16384,0.14449511633978948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,16384,0.09727378023995294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,65536,0.3184364371829563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,65536,0.5545146730211046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,65536,0.37552621629503036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,12288,0.1078631083170573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,12288,0.07390222284528944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,10240,0.09185244639714558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,10240,0.062446223364935986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,8192,0.04344888859324985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,8192,0.07378044393327501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,8192,0.05068177647060818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,12288,0.06306755542755127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,10240,0.05412266651789347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,7168,0.0649075574345059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,7168,0.04549955659442478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,7168,0.03935111231274075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,6144,0.05508977837032742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,6144,0.0399253335263994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,5120,0.02982577681541443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,5120,0.046872887346479625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,5120,0.034753776258892484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,4096,0.03703377644220988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,4096,0.02940355406867133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,3584,0.03314400050375197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,3584,0.02663822140958574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,6144,0.033960888783137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,4096,0.02594311038653056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,3584,0.023758222659428913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,3072,0.02090399960676829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,3072,0.0284551117155287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,3072,0.02409511142306858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,2560,0.024900444679790076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,2560,0.021317332983016968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,2560,0.018775110443433125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,2048,0.016578666037983365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,2048,0.020304888486862183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,2048,0.01861422260602315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,1536,0.01622577839427524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,1536,0.016047999262809753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,1024,0.01238844460911221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,1024,0.012244444754388599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,1536,0.014534221755133735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,1024,0.013377777404255338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,768,0.010416000253624385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,768,0.011769777370823754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,512,0.010621333287821876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,512,0.008454221818182204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,512,0.00979644474056032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,768,0.011505777637163797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,256,0.005971555494599872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,256,0.008936000367005667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,256,0.008779555559158325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,128,0.005024888863166173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,32,0.004722666823201709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,128,0.0081377774477005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,6144,64,0.004492444296677907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,64,0.008797333472304875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,6144,32,0.008969777988062965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,65536,0.4777830971611871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,65536,0.3708808951907688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,6144,128,0.007800000409285228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,16384,0.07217422458860609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,65536,0.2833991050720215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,16384,0.12110755178663467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,16384,0.09679288996590509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,12288,0.09273866812388103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,12288,0.07416355609893799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,10240,0.0758062203725179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,10240,0.062104887432522245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,12288,0.05487199955516391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,10240,0.047510223256217114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,8192,0.03881866733233134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,8192,0.06220622195137871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,8192,0.05105333195792305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,7168,0.053752889235814415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,7168,0.045167111688190036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,7168,0.03510844376352098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,6144,0.04712355468008253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,6144,0.04011199871699015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,5120,0.026729777455329895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,5120,0.0402124457889133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,5120,0.0347306662135654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,6144,0.030541333887312148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,4096,0.03238222334120009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,4096,0.029299557209014893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,3584,0.028934223784340754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,3584,0.026693332526418898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,3072,0.018790221876568265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,3072,0.025012445118692186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,3072,0.02395733363098568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,4096,0.023100444012218054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,2560,0.016917333006858826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,3584,0.02137600051032172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,2560,0.021574222379260596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,2560,0.021342222889264423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,2048,0.015051555302408008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,2048,0.018174222773975797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,2048,0.018594667315483093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,1536,0.01498133275243971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,1536,0.01586666703224182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,1024,0.011337777806652917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,1024,0.011687999798191918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,1536,0.013305777476893531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,1024,0.012864888542228274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,768,0.010002666877375709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,768,0.011049778097205691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,512,0.007983999947706858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,768,0.01051644484202067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,512,0.009672889278994666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,256,0.00609777785009808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,512,0.009513777991135916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,256,0.008799110849698385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,128,0.007139555282062954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,128,0.0053119998839166425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,5120,256,0.007988444632954067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,128,0.008115555677149031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,64,0.004380444271696938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,64,0.008685333033402761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,5120,32,0.004569777597983678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,5120,32,0.008719111482302347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,16384,0.061384889814588756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,65536,0.3730782137976752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,65536,0.2248328791724311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,16384,0.09507555431789821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,16384,0.05635466840532091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,12288,0.07081244389216106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,12288,0.04435733291837904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,10240,0.041004445817735456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,10240,0.06064089139302572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,10240,0.038422223594453596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,65536,0.2398142284817166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,8192,0.04798755711979336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,8192,0.03203199969397651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,12288,0.047741333643595375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,7168,0.04313866628540886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,8192,0.033371554480658636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,7168,0.0303182237678104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,7168,0.029041777054468792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,6144,0.03661511010593838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,6144,0.02607022225856781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,5120,0.02366399930583106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,5120,0.030761778354644775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,5120,0.023017777336968318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,4096,0.025892444782786902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,4096,0.019914666811625164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,6144,0.026687999566396076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,4096,0.020440000626775954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,3584,0.018969777557584975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,3584,0.022601778308550518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,3584,0.018358222312397428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,3072,0.016695110334290397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,3072,0.019965334071053397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,3072,0.016880000631014507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,2560,0.01711466742886437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,2560,0.015446222490734525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,2048,0.013543999857372709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,2560,0.015234667393896313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,2048,0.014361777239375643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,2048,0.013959111438857185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,1536,0.012103111379676394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,1536,0.011920000116030375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,1536,0.012265777422322167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,1024,0.010452444354693094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,1024,0.00925155563486947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,1024,0.01056177748574151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,768,0.007770666645632849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,768,0.008928889201747047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,512,0.006154666758245892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,512,0.007977777885066139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,256,0.004958222309748332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,256,0.007378666765160031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,128,0.004337777694066365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,128,0.006883555402358373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,64,0.004562666846646203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,64,0.007294221884674496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,768,0.00978488889005449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,256,0.006903111106819577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,512,0.008538666698667739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,4096,128,0.006082666830884085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,4096,32,0.004736888739797804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,4096,32,0.0074026667409473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,16384,0.0830471118291219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,65536,0.34429332945081925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,65536,0.21059021684858534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,16384,0.05534933341874016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,12288,0.050395554966396756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,12288,0.061965333090888135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,16384,0.06708444489373101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,65536,0.25461334652370876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,12288,0.04320266511705187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,10240,0.05210933420393202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,10240,0.03733511103524102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,8192,0.03563022282388475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,8192,0.042083554797702365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,8192,0.031189332405726116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,10240,0.04294844468434652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,7168,0.032362666394975446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,7168,0.03752977649370829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,7168,0.02826311190923055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,6144,0.028008888165156048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,6144,0.03305244445800781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,6144,0.025637333591779072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,5120,0.024792000651359558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,5120,0.027775999572541978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,5120,0.022760889596409265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,4096,0.021362667282422382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,4096,0.022408889399634466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,4096,0.01960000064637926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,3584,0.019713777634832595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,3584,0.019911999503771465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,3584,0.018187556001875136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,3072,0.017275555266274344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,3072,0.01774222155412038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,3072,0.01683644453684489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,2560,0.015828443898095023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,2560,0.015443555182880826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,2560,0.015289778510729471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,2048,0.013001778059535556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,2048,0.013721778160995908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,2048,0.013950222068362765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,1536,0.010803555448849997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,1536,0.012316444681750404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,1024,0.008551110823949179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,1024,0.009687110781669617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,768,0.007050666544172499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,768,0.008811555802822113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,512,0.005778666585683823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,512,0.00792711145348019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,256,0.0046817776229646476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,256,0.007356444166766272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,128,0.006239111224810283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,128,0.004138666722509596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,1536,0.012285333540704517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,1024,0.010861333045694562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,128,0.006868444383144379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,64,0.003951999876234266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,768,0.009415999882751042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,512,0.008072888685597314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3584,256,0.006688888702127669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,64,0.007181333170996771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3584,32,0.003968888686762916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3584,32,0.007322666545708974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,65536,0.29235021273295086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,65536,0.20384532875484893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,65536,0.22307289971245658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,16384,0.07860089010662503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,16384,0.05377422107590569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,12288,0.05612888601091173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,12288,0.04193777839342753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,10240,0.03956088754865859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,10240,0.04670222269164192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,10240,0.036722666687435575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,8192,0.03808800048298306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,8192,0.030586666531032983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,7168,0.02959288822280036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,7168,0.03338311115900675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,16384,0.059502224127451576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,7168,0.027959111664030287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,6144,0.025959110922283594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,6144,0.02854666776127285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,12288,0.046037332879172436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,6144,0.025031111306614343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,5120,0.023141332798533972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,5120,0.024023110667864483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,5120,0.022566222482257422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,8192,0.032960888412263654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,4096,0.019910222954220243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,4096,0.019523556033770244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,3584,0.01792888840039571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,3584,0.018194667167133756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,3072,0.015632000234391954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,3072,0.016522667474216886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,4096,0.019903111788961623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,3584,0.018622222873899672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,3072,0.016287111573749118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,2560,0.01370488852262497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,2560,0.015184889237085978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,2048,0.011643555429246692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,2560,0.014843554960356818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,2048,0.013624000052611033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,2048,0.013208889298968844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,1536,0.011776000261306763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,1536,0.009764444496896531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,1536,0.011991110940774282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,1024,0.01019377758105596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,1024,0.00777511133088006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,1024,0.009723555710580613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,768,0.009010666774378883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,768,0.006518222391605377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,768,0.008584000170230865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,512,0.005399999933110342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,512,0.007912889122962952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,256,0.004450666821665234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,256,0.0073315559162033936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,512,0.007171555525726742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,256,0.006163555714819167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,3072,128,0.00564444437623024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,128,0.003930666794379552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,128,0.006864000111818314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,64,0.003588444242874781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,64,0.007086222370465596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,3072,32,0.0037786666717794207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,3072,32,0.007215110792054071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,65536,0.19658845000796846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,65536,0.2383333312140571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,65536,0.19959644476572672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,16384,0.06323022312588163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,16384,0.05322666631804573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,12288,0.042618665430280894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,12288,0.046774221791161426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,12288,0.041688889265060425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,8192,0.0326533317565918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,10240,0.0396364430586497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,10240,0.036425779263178505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,16384,0.055220444997151695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,10240,0.03657955593532986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,8192,0.030454221698972914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,7168,0.028951999213960435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,8192,0.030163556337356567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,7168,0.027474666635195415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,7168,0.027216888136333887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,6144,0.024018666810459558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,6144,0.024861334098709956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,6144,0.025050666597154405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,5120,0.021372444099850122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,5120,0.021213332811991375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,5120,0.022322666313913133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,4096,0.017383111847771537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,4096,0.01851377718978458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,4096,0.019314666589101154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,3584,0.017317333155208163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,3584,0.015726221932305228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,3584,0.017858665850427415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,3072,0.01519377695189582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,3072,0.013996443814701505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,3072,0.016435555285877652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,2560,0.012220444778601328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,2560,0.015102222561836243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,2560,0.013800000150998434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,2048,0.010446222292052375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,2048,0.012404444317022959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,2048,0.013302221894264221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,1536,0.008882666627566019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,1536,0.011571555501884885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,1024,0.009249777429633671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,1024,0.006883555402358373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,1024,0.009272000028027428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,768,0.007982222570313348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,1536,0.010982222027248807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,768,0.005938666562239329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,768,0.008565333154466417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,512,0.005105777747101254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,512,0.007819555699825287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,256,0.005858666780922148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,256,0.004280888785918553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,256,0.007286222444640265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,128,0.005487999982304043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2560,512,0.0070364442136552595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,128,0.003875555677546395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,128,0.006780444333950679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,64,0.003527111063400904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,64,0.006968888971540663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2560,32,0.003710222327046924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2560,32,0.007072889142566257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,65536,0.16925599839952257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,65536,0.18901688522762725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,16384,0.05271200007862515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,65536,0.19436800479888916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,16384,0.05283555719587538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,12288,0.041625777880350746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,12288,0.04153777824507819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,10240,0.030584888325797185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,10240,0.03410311208830939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,10240,0.03613155417972141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,8192,0.02780888809098138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,16384,0.04552622305022346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,8192,0.030031998952229817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,12288,0.03504177927970886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,7168,0.023535999986860488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,8192,0.025711110896534387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,5120,0.018079999420377944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,7168,0.027709333433045283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,7168,0.02479644450876448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,6144,0.020612445142534044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,6144,0.02127200033929613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,6144,0.024744888146718342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,5120,0.02200711104604933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,4096,0.016255110502243042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,4096,0.01752355529202355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,4096,0.018920888503392536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,5120,0.018419555491871305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,3584,0.014785778191354541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,3584,0.017650667164060805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,3072,0.012281777958075205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,3072,0.016215110818545025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,3584,0.015478221906556023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,2560,0.012503999802801343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,2560,0.011167111496130625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,2560,0.014871110518773397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,2048,0.011277333729796939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,3072,0.013409777647919126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,2048,0.009319999979601966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,2048,0.012816888590653738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,1536,0.008204444415039485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,1536,0.010816000401973724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,1024,0.006425777657164468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,1024,0.00905777762333552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,1536,0.010063110954231685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,768,0.005656888915432825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,768,0.008616000413894653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,512,0.006364444477690591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,512,0.0047644442982143825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,512,0.007753777835104201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,1024,0.007727999654081132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,256,0.005642666584915585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,256,0.004129777765936322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,256,0.007163555257850223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,768,0.0069884442620807225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,128,0.0036728887094391715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,2048,128,0.005256888767083486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,128,0.006690666493442323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,64,0.0033413333197434745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,64,0.006883555402358373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,2048,32,0.0035439998739295532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,2048,32,0.0069288888739215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,65536,0.15127644273969862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,65536,0.19344622559017607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,65536,0.1728853384653727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,16384,0.0418124430709415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,16384,0.052170667383405894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,12288,0.036552889479531184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,16384,0.048250665267308555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,12288,0.032609777318106756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,12288,0.041125333971447416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,10240,0.03151288959715102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,10240,0.028472890456517536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,10240,0.03571822245915731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,8192,0.026203556193245783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,8192,0.028020444843504164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,8192,0.02996622191535102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,7168,0.020605333977275424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,7168,0.02721600068940057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,6144,0.021054221524132624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,6144,0.017343110508388944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,6144,0.024687111377716064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,7168,0.023989333046807185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,5120,0.015627556376987033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,5120,0.02190133266978794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,4096,0.014476444986131458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,4096,0.01885511146651374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,3584,0.012664000193277994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,3584,0.017532444662517972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,3072,0.010464000205198923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,3072,0.01614488826857673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,2560,0.01218488895230823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,2560,0.009479111267460717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,2560,0.014700444208251106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,2048,0.010829333629873065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,5120,0.01872977779971229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,3584,0.01534222231970893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,4096,0.01646399994691213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,3072,0.013454222016864352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,2048,0.008185777399275038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,2048,0.012297777665985955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,1536,0.007095110913117726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,1536,0.010453333457310995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,1024,0.005403555515739653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,1024,0.008943110704421997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,768,0.006631111105283101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,768,0.004824888788991504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,1536,0.0087333329849773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,1024,0.007076444725195567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,768,0.008394666843944127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,512,0.006088888893524806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,512,0.004265777766704559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,512,0.007717333734035492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,256,0.00544799988468488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,256,0.0037288889288902283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,256,0.007165333463086023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1536,128,0.00509688879052798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,128,0.0034017778105205963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,128,0.006720888945791457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,64,0.0032702222880389956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,64,0.006831999868154526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1536,32,0.003334222154484855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1536,32,0.00684533350997501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,65536,0.0993422203593784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,65536,0.19167199399736193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,65536,0.12627732753753662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,16384,0.03000266684426202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,16384,0.05187466740608215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,12288,0.027253333065244887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,12288,0.02454400062561035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,12288,0.04086222251256307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,16384,0.03528977764977349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,10240,0.02197422252760993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,10240,0.03566755519972907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,10240,0.024001777172088623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,8192,0.020282665888468426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,8192,0.019592000378502738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,8192,0.02964799933963352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,7168,0.01867644488811493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,7168,0.017456000049908955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,7168,0.027143999934196472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,6144,0.016543999314308167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,6144,0.013808888693650564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,6144,0.0243857784403695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,5120,0.014883556299739413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,5120,0.012154666913880242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,5120,0.021715554926130507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,4096,0.011108444796668159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,4096,0.018757333358128864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,3584,0.0125937776433097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,3584,0.009871111147933537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,3584,0.017328888177871704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,4096,0.013266666895813413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,3072,0.009333333207501305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,3072,0.01590399940808614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,2560,0.007888889147175683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,3072,0.010984888507260216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,2560,0.013817778064144982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,2048,0.008616000413894653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,2048,0.006802666518423293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,2048,0.011791999969217511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,1536,0.007502222226725684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,2560,0.010093332992659675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,1536,0.0058168888919883305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,1536,0.010420444111029306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,1024,0.006572444405820634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,1024,0.004872888740566042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,1024,0.008836444053384993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,768,0.004527111020353106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,768,0.008247999681366814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,512,0.005686222265164058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,512,0.004041777716742622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,768,0.006169777777459886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,512,0.007607999775144789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,256,0.00517866677708096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,256,0.003563555578390757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,256,0.007098666495747036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,1024,128,0.004920888692140579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,128,0.003306666596068276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,128,0.00666933341158761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,64,0.003089777711364958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,64,0.006814222368929122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,1024,32,0.0032239999208185407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,1024,32,0.006762666834725275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,65536,0.07833777533637153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,16384,0.024433778391944036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,65536,0.1916337807973226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,16384,0.051538666089375816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,12288,0.026687999566396076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,12288,0.020036444067955017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,65536,0.119687106874254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,12288,0.04097155398792691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,16384,0.03440266516473558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,10240,0.01835466590192583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,10240,0.03540088733037313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,10240,0.023584889041052923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,8192,0.017479999197853934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,8192,0.029719998439153034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,8192,0.019885333047972787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,7168,0.018583999739752877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,7168,0.015684444043371413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,7168,0.026920000712076824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,6144,0.016167110866970487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,6144,0.013596444494194455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,6144,0.02439022229777442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,5120,0.01444355481200748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,5120,0.011973333027627734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,5120,0.021669333179791767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,4096,0.010294222169452244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,4096,0.018631999691327412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,3584,0.011964444484975604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,3584,0.009571555587980483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,3584,0.01737155516942342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,3072,0.008806222014957005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,3072,0.015581332974963717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,4096,0.012900444368521372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,3072,0.010208889014191097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,2560,0.007999999655617608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,2560,0.013564444250530668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,2560,0.009293333523803288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,2048,0.0070222218831380205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,2048,0.01182755579551061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,1536,0.007497777541478475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,1536,0.005695111221737332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,2048,0.008295111358165741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,1536,0.010086221827401055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,1024,0.00646577775478363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,1024,0.004801777915822135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,1024,0.00887822194231881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,768,0.006036444256703059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,768,0.004361777669853634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,768,0.00832444429397583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,512,0.0038782221575578055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,512,0.005600000007285013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,512,0.007660444411966536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,256,0.005125333451562458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,256,0.0034764446318149567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,256,0.007064888874689738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,128,0.0031911111954185697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,128,0.00666133314371109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,64,0.003039111072818438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,64,0.006798222247097228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,768,32,0.0030871110243929755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,768,32,0.006800000038411882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,65536,0.054678221543629967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,16384,0.019367999500698514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,65536,0.1926577753490872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,16384,0.051510221428341336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,12288,0.024503111839294434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,12288,0.015601777368121676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,12288,0.040848887628979154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,10240,0.014049778381983439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,10240,0.03542222248183356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,8192,0.018417777286635507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,8192,0.015140444040298462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,8192,0.029521776570214167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,768,128,0.004874666531880696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,7168,0.017086222767829895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,16384,0.03157688842879401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,65536,0.10852266682518853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,10240,0.021751110752423603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,7168,0.026897778113683064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,7168,0.01275022245115704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,6144,0.015055111712879606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,6144,0.012248888611793518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,6144,0.02420622275935279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,5120,0.013627555635240344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,5120,0.011023999916182624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,5120,0.021563554803530376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,4096,0.010099555883142684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,4096,0.018534221582942538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,3584,0.008831111093362173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,3584,0.01715288890732659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,4096,0.012062222593360476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,3072,0.010108444425794814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,3072,0.007978666987684038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,3072,0.015030221806632148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,2560,0.008547555241319869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,3584,0.011177777416176267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,2560,0.007683555285135905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,2560,0.013074666261672974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,2048,0.006771555377377405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,2048,0.011360000405046674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,1536,0.005719111197524601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,1536,0.010217777556843227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,2048,0.007689778175618913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,1536,0.006897777732875612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,1024,0.0047848886913723415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,1024,0.008800889054934183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,768,0.005747555444637935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,768,0.004320000194840961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,1024,0.006175111151403851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,768,0.008199111455016667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,512,0.003847111016511917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,512,0.007542221910423703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,512,0.005407111098368962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,256,0.003391111062632667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,256,0.00702044450574451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,128,0.0046880000995265115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,128,0.003113777687152227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,128,0.006654222392373615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,64,0.002903999967707528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,512,256,0.004957333207130432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,64,0.006744888921578725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,512,32,0.0030053332448005676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,512,32,0.006741333338949416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,65536,0.045979556110170156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,65536,0.19387111398908827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,16384,0.014458666245142618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,16384,0.052024887667761914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,12288,0.012301333248615265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,12288,0.04080266753832499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,10240,0.01089244418674045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,10240,0.035240888595581055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,8192,0.01752711170249515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,8192,0.010850666297806634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,65536,0.10763911406199138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,16384,0.029859556092156306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,12288,0.0236817780468199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,10240,0.02088977727625105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,8192,0.029792000850041706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,7168,0.016322667400042217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,7168,0.010447111394670276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,7168,0.02700800034734938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,6144,0.014479110638300577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,6144,0.009733333355850643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,6144,0.024263999528355066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,5120,0.00925333301226298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,5120,0.021520889467663232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,4096,0.008307555483447181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,4096,0.017829333742459614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,3584,0.008112000094519721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,5120,0.012889777620633444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,4096,0.010913777682516309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,3584,0.01603200038274129
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,3072,0.00755466686354743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,3072,0.014204444156752693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,3584,0.010065777434243096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,2560,0.006500444478458828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,2560,0.012720888687504662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,3072,0.009033777647548253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,2560,0.008231999973456064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,2048,0.005809777726729711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,2048,0.007455999652544658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,2048,0.011388444238238864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,1536,0.006688888702127669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,1536,0.005048888838953442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,1536,0.010055999788973067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,1024,0.00433688900536961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,1024,0.008728000024954477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,768,0.003979555434650845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,768,0.008175111479229396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,512,0.003637333297067218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,1024,0.005961777849329843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,768,0.005623999983072281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,512,0.005201777650250329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,512,0.0075333333677715724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,256,0.00479911102188958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,256,0.0032088889016045462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,256,0.0069431112044387394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,256,128,0.004592888885074192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,128,0.0029537777105967202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,128,0.006602666858169768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,64,0.0028088889602157804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,64,0.006637333167923822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,256,32,0.0028284444577164124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,256,32,0.006650666809744305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,65536,0.10774399836858113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,65536,0.033815999825795494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,65536,0.19360444280836317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,16384,0.01184977756606208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,16384,0.052150222990247935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,12288,0.02424000038041009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,12288,0.010503999888896942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,12288,0.040680888626310564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,10240,0.02113155523935954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,10240,0.009986666341622671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,10240,0.03539199961556329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,8192,0.010114666488435533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,8192,0.0295724438296424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,7168,0.01646311084429423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,16384,0.030583110120561387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,7168,0.009675555759006077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,7168,0.026741334133678015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,6144,0.014448889427714877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,6144,0.008984000318580205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,6144,0.024199111594094172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,5120,0.012392000191741519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,5120,0.008055999875068665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,5120,0.021427555216683283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,4096,0.0107004443804423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,8192,0.01789066692193349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,4096,0.007400000260935889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,4096,0.017482666505707633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,3584,0.007298666569921706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,3584,0.015563555889659457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,3072,0.007048889166778988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,3072,0.013988445202509562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,2560,0.008212444682916006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,3584,0.009967111051082611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,2560,0.0063484443558586975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,2560,0.012611555556456247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,2048,0.005597333527273602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,3072,0.00922400007645289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,2048,0.011317333413494958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,1536,0.006711111300521427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,1536,0.004877333425813251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,1536,0.00998933364947637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,1024,0.005939555664857228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,2048,0.007419555551475949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,1024,0.004320888883537716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,1024,0.008741333252853816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,768,0.005631111148330901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,768,0.004033777862787247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,768,0.00816533300611708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,512,0.003677333394686381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,512,0.0074746666683091064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,256,0.003253333270549774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,256,0.0069288888739215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,128,0.0029768887907266617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,512,0.005190222213665645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,256,0.0047893333766195504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,128,0.006573333508438534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,64,0.002846222163902389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,64,0.006638222270541721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,128,32,0.003050666716363695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,128,32,0.006585777633719974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,512,128,128,0.0046160001721647055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,65536,0.027519999278916255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,16384,0.010972444381978778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,65536,0.19370222091674805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,16384,0.05202844407823351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,12288,0.009749333063761393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,12288,0.04054222173160977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,10240,0.009296000003814697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,10240,0.035176889763938055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,8192,0.008816888762844933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,8192,0.02935289012061225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,7168,0.00833777752187517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,7168,0.026755554808510676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,6144,0.0074106670088238185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,6144,0.023994666006830003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,5120,0.007373332977294922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,5120,0.02103288968404134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,4096,0.006894222150246303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,4096,0.017126222451527912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,3584,0.007121777368916406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,3584,0.01533600025706821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,3072,0.006782222125265334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,3072,0.013944889108339945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,2560,0.006158222340875202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,2560,0.012657778130637275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,2048,0.005514666852023866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,2048,0.01128000020980835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,1536,0.004794666750563515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,1536,0.009956444303194681
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,1024,0.004105777790149053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,1024,0.00871022211180793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,768,0.0037502224246660867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,768,0.0081395556529363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,512,0.003415999934077263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,512,0.0074142225914531285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,256,0.003060444361633725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,256,0.006879999819729064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,128,0.002862222285734283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,128,0.006488889041874144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,64,0.0026702222724755607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,64,0.006600889066855113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,64,32,0.0027280000762807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,64,32,0.006565333240562015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,65536,0.025655999779701233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,65536,0.19545511404673258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,16384,0.010289777484205034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,16384,0.052038222551345825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,12288,0.009114666945404476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,12288,0.04042666819360521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,10240,0.008757332960764566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,10240,0.03505333264668783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,8192,0.008976000050703684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,8192,0.0295448899269104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,7168,0.007828444242477417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,7168,0.02676800058947669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,6144,0.007150222029950883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,6144,0.023927999867333308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,5120,0.007129777636792924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,5120,0.020827554994159274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,4096,0.006751110984219446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,4096,0.01685422162214915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,3584,0.007128888534175024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,3584,0.01531733406914605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,3072,0.006616000086069107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,3072,0.013933334085676404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,2560,0.006150222072998683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,2560,0.012544000314341651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,2048,0.005487999982304043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,2048,0.011262222296661802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,1536,0.0047848886913723415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,1536,0.009950222240553962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,1024,0.00408000002304713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,1024,0.008725333544943068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,128,0.0028195555011431375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,768,0.0037600000699361167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,768,0.008153777983453538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,512,0.0033982222278912864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,512,0.0074622225430276655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,256,0.003032888803217146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,256,0.00685511115524504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,128,0.006528000036875407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,64,0.002667555585503578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,64,0.006559999866618051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,512,32,32,0.0026764445420768526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,512,32,32,0.006578666468461354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,16384,1.2651964823404949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,16384,0.5969457626342773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,16384,0.6266586515638564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,12288,0.9196853637695312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,12288,0.4518631299336751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,12288,0.45616886350843644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,10240,0.7467759980095757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,8192,0.5781982209947374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,10240,0.3718142244550917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,8192,0.29798756705390084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,7168,0.5090471373664008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,7168,0.26570312182108563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,6144,0.4293999936845567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,6144,0.2326400015089247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,5120,0.19603821966383192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,5120,0.36676711506313747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,6144,0.22243022918701172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,5120,0.18825777371724448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,4096,0.15789688958062068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,4096,0.2938720120324029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,4096,0.15501599841647679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,3584,0.25406132804022896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,3584,0.13707110616895887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,3072,0.12303911315070258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,3072,0.2241591082678901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,3072,0.11779556009504531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,2560,0.18787288665771484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,2560,0.10303644339243571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,2048,0.0856142242749532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,2048,0.1546471118927002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,2048,0.08400711086061265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,1536,0.06964355707168579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,1536,0.11813510788811578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,1536,0.06610221995247735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,1024,0.08272622028986613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,1024,0.05044088761011759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,768,0.043136000633239746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,768,0.06565688716040717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,768,0.04303822252485487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,512,0.03451555636194017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,512,0.04756977823045519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,10240,0.3865777651468913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,512,0.03818577859136794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,8192,0.30362311999003094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,256,0.027815110153622095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,256,0.03359377715322707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,7168,0.27177156342400444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,3584,0.13875467247433132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,2560,0.10521066851086086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,256,0.03229511115286086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,128,0.023802666200531855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,32,0.02188266648186578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,128,0.024944000773959692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,65536,1024,0.05053155620892843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,128,0.029912001556820337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,65536,64,0.02330933345688714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,64,0.03142933381928338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,65536,32,0.033928887711630926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,65536,1.1521120071411133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,16384,0.15498488479190403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,16384,0.29437867800394696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,12288,0.11674400170644124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,16384,0.15390400091807047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,12288,0.21878311369154188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,12288,0.11840266651577419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,65536,0.6296515464782715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,10240,0.0981973343425327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,10240,0.18572533130645752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,65536,0.6176844702826606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,10240,0.09571911229027642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,8192,0.0815991097026401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,8192,0.1472231149673462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,8192,0.08050488763385348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,7168,0.1261733373006185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,7168,0.07134399811426799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,7168,0.07164888911777072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,6144,0.10875733031166925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,6144,0.061400002903408475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,5120,0.052890668312708534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,5120,0.09293422434065078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,6144,0.06147199869155884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,5120,0.05102133419778612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,4096,0.07650578022003174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,4096,0.04185333185725742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,3584,0.06637599733140734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,4096,0.04312444395489163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,3584,0.03809422254562378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,3072,0.056326223744286426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,3584,0.039268443981806435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,3072,0.034104890293545194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,2560,0.02999911043379042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,3072,0.03432710965474447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,2560,0.04850577645831638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,2560,0.029599110285441082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,2048,0.025645333859655593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,2048,0.03900977638032701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,2048,0.025235555238193933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,1536,0.030187557140986126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,1536,0.020949333906173706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,1024,0.017816889617178176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,1024,0.021888888544506494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,1536,0.02184177769554986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,1024,0.017157332764731515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,768,0.015949333707491558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,768,0.01763199932045407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,768,0.015375110838148328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,512,0.013853333062595792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,512,0.013653332988421122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,512,0.01365244471364551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,256,0.010107555323176915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,256,0.011491555306646558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,128,0.007556444240940942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,128,0.00980000032318963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,256,0.011729777687125735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,64,0.0064151109092765385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,16384,128,0.010024888647927178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,64,0.010563555690977307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,16384,32,0.006812444577614467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,16384,32,0.011391110718250275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,16384,0.14430310991075304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,65536,0.55549865298801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,65536,0.817661338382297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,65536,0.5829502211676704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,16384,0.2102053297890557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,16384,0.14078844918145075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,12288,0.15464977423350015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,12288,0.1084773341814677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,12288,0.11039111349317764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,10240,0.0914728906419542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,10240,0.13209954897562662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,10240,0.0950088898340861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,8192,0.07472178008821276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,8192,0.10642577542199029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,8192,0.07458844449785021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,7168,0.06581155459086101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,7168,0.09352978070576985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,7168,0.0653288894229465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,6144,0.08204533656438191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,6144,0.05735199981265598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,5120,0.04884799983766344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,5120,0.06674666537178887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,6144,0.05705510907702976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,5120,0.04894222153557671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,4096,0.04009955459170871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,4096,0.05361244413587782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,4096,0.04024444354905023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,3584,0.048251556025611035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,3584,0.03602666656176249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,3072,0.0322017769018809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,3584,0.03594755464129978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,3072,0.04194577866130405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,3072,0.03211377726660834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,2560,0.03616177704599168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,2560,0.028738667567571003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,2560,0.028128001425001357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,2048,0.024736000431908503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,2048,0.029151999288135107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,2048,0.024178667200936213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,1536,0.021249777740902368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,1536,0.02299733294381036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,1536,0.020071110791630216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,1024,0.01683644453684489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,1024,0.016954667038387723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,768,0.015368888775507608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,768,0.014116444521480136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,768,0.015112888481881885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,512,0.011343999869293638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,512,0.01353422221210268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,128,0.00923911150958803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,256,0.008360000120268928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,1024,0.01726222203837501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,256,0.010527110762066312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,128,0.006333333336644703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,128,0.009777777724795872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,64,0.005537777725193236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,64,0.01051288925939136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,12288,32,0.005931555396980709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,12288,32,0.010769777827792697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,65536,0.8183662096659342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,512,0.01294222225745519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,65536,0.7648737695482043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,65536,0.4001715448167589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,12288,256,0.010657777388890585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,16384,0.1781431171629164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,16384,0.10131200154622395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,12288,0.1386044422785441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,12288,0.1557368834813436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,12288,0.07811644342210558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,10240,0.1297644509209527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,16384,0.214791112475925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,10240,0.11490221818288167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,10240,0.06439644760555692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,8192,0.10774755477905273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,8192,0.09420533312691583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,8192,0.05317599905861748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,7168,0.09529155492782593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,7168,0.08157777786254883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,7168,0.04730311036109924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,6144,0.06873422198825412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,6144,0.04181155562400818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,6144,0.08038933409584893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,5120,0.06826044453514947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,5120,0.059009777175055615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,5120,0.035473777188195124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,4096,0.04808177881770664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,4096,0.02985066506597731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,3584,0.042338665988710195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,3584,0.027288888891537983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,3072,0.0365973346763187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,3072,0.024516445067193773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,2560,0.03310666812790765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,2560,0.03156000044610765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,2560,0.02156266735659705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,2048,0.027440889014138117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,2048,0.025954667064878676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,2048,0.01882844501071506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,1536,0.020555555820465088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,1536,0.016163556112183463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,4096,0.05466844307051765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,3584,0.04664177695910136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,3072,0.04154755671819051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,1024,0.01532444523440467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,1024,0.013597333596812354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,768,0.012792888614866467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,768,0.012438221938080259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,512,0.012710221939616732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,512,0.010293333066834344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,512,0.011006222003036074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,256,0.010536889235178629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,256,0.007352888584136963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,256,0.008736000292830998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,1536,0.02158311174975501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,1024,0.017102221647898357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,128,0.009138666921191746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,128,0.0058240000572469496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,128,0.008147555920812819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,64,0.005623111294375525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,10240,768,0.015306666493415833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,64,0.008784888519181145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,10240,32,0.005777777896987067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,10240,32,0.009287111461162567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,65536,0.5842817624409994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,65536,0.37412354681226945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,16384,0.14231733481089273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,12288,0.06285066737069024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,16384,0.09654044442706639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,12288,0.10857066843244766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,12288,0.07577955722808838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,10240,0.09047289027108087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,10240,0.062182221147749156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,8192,0.07365333371692233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,8192,0.05057511064741346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,7168,0.0653431084420946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,16384,0.08348888821072048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,65536,0.31582487954033744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,8192,0.043015112479527794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,10240,0.05296444561746386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,7168,0.04528977804713779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,6144,0.056157330671946205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,6144,0.03981155488226149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,5120,0.04719110992219713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,5120,0.034719109535217285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,4096,0.03839288817511664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,4096,0.029136889510684546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,3584,0.022882666852739122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,3584,0.02646133303642273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,3584,0.033824000093672014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,7168,0.038226667377683855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,5120,0.029520889123280842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,6144,0.03427199853791131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,3072,0.02089244458410475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,3072,0.029324445459577773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,3072,0.023852444357342188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,4096,0.024894222617149353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,2560,0.024736000431908503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,2560,0.021225778592957392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,2048,0.016385777129067313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,2560,0.018710222509172227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,2048,0.020818667279349435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,2048,0.01830311119556427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,1536,0.014411555396185981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,1536,0.01641244524055057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,1536,0.015951111912727356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,1024,0.012270222107569376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,1024,0.012433778080675336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,1024,0.013559111290507846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,768,0.011336888704035016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,768,0.010566222170988718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,768,0.012301333248615265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,512,0.008555555509196388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,512,0.010028444230556488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,256,0.006165333506133821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,256,0.008804444637563493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,512,0.010423111418883005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,256,0.008929777476522658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,128,0.005197333378924264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,128,0.008200888832410177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,64,0.004584889031118817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,8192,128,0.007687999970383114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,64,0.008599110775523717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,8192,32,0.004815111143721474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,8192,32,0.008854221966531541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,65536,0.4919226434495714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,65536,0.5008373260498047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,65536,0.36962578031751847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,16384,0.12654133637746176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,16384,0.09703822268380059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,12288,0.09490933020909627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,12288,0.0735128919283549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,16384,0.10694666703542073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,10240,0.07898310820261638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,10240,0.061863998572031655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,8192,0.054711110062069364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,12288,0.08099110921223958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,10240,0.06758399804433186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,8192,0.06417066521114774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,8192,0.050251556767357715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,7168,0.048560887575149536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,7168,0.05610133541954888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,7168,0.044975111881891884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,6144,0.04196888870663113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,6144,0.049626668294270836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,6144,0.03956800036960178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,5120,0.04225600096914503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,5120,0.03454666667514377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,4096,0.030327998929553564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,5120,0.03677155574162801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,4096,0.03383200036154853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,4096,0.02902488907178243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,3584,0.03010666701528761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,3584,0.026408000124825373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,3072,0.025016888976097107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,3072,0.025955556167496577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,3072,0.023759111762046814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,2560,0.02225422196918064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,2560,0.02255022194650438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,2560,0.021034666233592566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,3584,0.028008888165156048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,2048,0.019151111443837483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,2048,0.01865866614712609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,2048,0.0184879998366038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,1536,0.016773333152135212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,1536,0.015270221564504834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,1536,0.015792888071801927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,1024,0.012015110916561551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,1024,0.013464888764752282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,768,0.012447111308574677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,768,0.010303110712104374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,768,0.011584888729784222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,512,0.008634666601816813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,512,0.01001422190003925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,1024,0.014141332772043018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,256,0.008919999831252629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,256,0.00638133328821924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,256,0.00887822194231881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,128,0.00759733302725686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,128,0.005570666657553778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,128,0.00810933361450831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,7168,512,0.011034666664070554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,64,0.005049777941571342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,64,0.008615111311276754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,7168,32,0.0053688887920644544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,7168,32,0.00885688927438524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,16384,0.11175999376508926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,65536,0.29201867845323354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,65536,0.42745333247714573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,16384,0.09604177872339885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,12288,0.05716266896989611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,12288,0.08471733331680298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,65536,0.3710835509830051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,12288,0.07295999924341838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,10240,0.049413333336512245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,16384,0.07643644677268134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,10240,0.0702622201707628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,10240,0.061689780818091504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,8192,0.03973866502443949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,8192,0.056812445322672524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,8192,0.0507919987042745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,7168,0.036108444134394325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,7168,0.0488702224360572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,7168,0.04446488949987623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,6144,0.041695111327701144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,6144,0.0395359992980957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,5120,0.03610400027698941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,5120,0.03429244293106927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,4096,0.0236799998415841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,4096,0.0291884442170461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,4096,0.02889866630236308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,3584,0.021451556020312842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,3584,0.026132444540659588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,6144,0.03184088733461168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,5120,0.02807377775510152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,3584,0.026161778304311965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,3072,0.022615111536449854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,3072,0.023678221636348303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,2560,0.017804443836212158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,2560,0.019514666663275827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,2560,0.020995555652512446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,3072,0.019700444406933255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,2048,0.01607377827167511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,2048,0.01827022267712487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,2048,0.015771556231710646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,1536,0.013864888913101621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,1536,0.013088888592190213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,1536,0.01571822166442871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,1024,0.01018311083316803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,1024,0.01332800007528729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,768,0.010799999866220685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,768,0.008819555242856344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,768,0.010992888775136737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,512,0.009300444689061906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,512,0.007159999675220913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,512,0.00960355583164427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,256,0.00741155528359943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,256,0.005356444252861871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,256,0.008711111214425828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,1024,0.011865778101815118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,128,0.004551110996140374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,128,0.007995555798212687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,64,0.004039999925427967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,64,0.00852177788813909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,6144,32,0.004321777572234471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,6144,32,0.008608000146018134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,16384,0.11344710985819499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,16384,0.09457155730989243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,65536,0.36516621377733016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,16384,0.055016001065572105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,65536,0.21649332841237387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,65536,0.43602667914496523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,6144,128,0.006594666590293248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,12288,0.07933244440290663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,12288,0.07109689050250582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,12288,0.04280089007483589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,10240,0.059565332200792104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,10240,0.03728444377581278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,8192,0.04705955584843954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,8192,0.030990223089853924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,7168,0.04593955477078756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,7168,0.04234844446182251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,7168,0.028208888239330713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,10240,0.06559111012352838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,8192,0.05501066644986471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,6144,0.03630400035116408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,6144,0.04057777921358744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,6144,0.025273778372340735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,5120,0.031165334913465712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,5120,0.022594667143291895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,4096,0.02550044490231408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,4096,0.019490665859646268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,3584,0.02306488818592495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,5120,0.03464444478352865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,3584,0.018061333232455783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,4096,0.027595554788907368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,3072,0.019792889555295307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,3072,0.016680889659457736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,2560,0.0170160002178616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,3584,0.025034666061401367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,2560,0.015206222732861837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,1536,0.013926222920417786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,2048,0.016345777445369296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,2048,0.014351111319330005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,2048,0.013678222066826291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,1536,0.011956444217099084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,1536,0.012432888978057437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,1024,0.01162755572133594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,3072,0.021853332718213398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,2560,0.018534221582942538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,1024,0.009388444324334463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,1024,0.010272000398900773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,768,0.010598222414652506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,768,0.008211555580298105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,256,0.007338666253619724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,768,0.00884622252649731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,512,0.006689777804745569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,512,0.008799110849698385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,512,0.00793333351612091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,64,0.004184000194072723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,256,0.007279111279381647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,256,0.004947555561860402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,128,0.004257777912749184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,128,0.0069022224181228215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,64,0.00721244431204266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,5120,32,0.004338666796684265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,5120,32,0.007430222299363878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,65536,0.21490577856699625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,5120,128,0.006495999793211619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,65536,0.2826640076107449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,16384,0.07936000161700778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,16384,0.0522346662150489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,65536,0.19731644789377847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,12288,0.04371644390953911
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,12288,0.0588266650835673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,16384,0.05682133303748237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,12288,0.0410515566666921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,10240,0.048727999130884804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,10240,0.0357528891828325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,10240,0.037704888317320086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,8192,0.04196088843875461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,8192,0.03016266557905409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,7168,0.027997334798177082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,7168,0.0351511107550727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,7168,0.027496889233589172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,6144,0.025236444340811834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,6144,0.030809776650534734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,6144,0.02458577851454417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,5120,0.022348443667093914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,8192,0.031129777431488037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,5120,0.025996444953812495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,5120,0.02200444373819563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,4096,0.021891555852360193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,4096,0.01905955539809333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,3584,0.020434666011068556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,3584,0.017665778597195942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,3072,0.015758223003811307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,4096,0.018645332919226754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,3072,0.017112000121010672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,3584,0.017140444782045152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,3072,0.016387555334303115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,2560,0.013921777407328287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,2560,0.014847111370828418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,2048,0.012690666649076672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,2048,0.011843555503421359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,2048,0.013533333109484779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,1536,0.009841777384281158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,2560,0.014440000057220459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,1536,0.012105777859687805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,1024,0.00958933350112703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,1024,0.00793333351612091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,1536,0.011400000088744693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,1024,0.009463999834325578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,768,0.008447111480765873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,768,0.006789333290523953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,768,0.008817777865462834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,512,0.006948444578382704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,512,0.00535999983549118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,512,0.007857778006129796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,256,0.0058257778485616045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,256,0.004506666627195147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,256,0.007287999822033777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,4096,128,0.005327111317051782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,128,0.004013333469629288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,128,0.006794666664467917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,64,0.003576888806290097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,64,0.007113777928882175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,4096,32,0.003818666769398583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,4096,32,0.007256888680987888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,65536,0.2624239921569824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,65536,0.20747555626763237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,65536,0.19152888986799452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,16384,0.06982488764656915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,16384,0.05141244331995646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,16384,0.05574755536185371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,12288,0.055649777253468834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,12288,0.040651556518342756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,10240,0.037307557132509016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,10240,0.04797600044144524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,10240,0.03552088803715176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,8192,0.03071911136309306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,12288,0.04311555624008179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,8192,0.03646488984425863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,8192,0.029828445778952703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,7168,0.03270666797955831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,7168,0.02717511190308465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,6144,0.024999111890792847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,6144,0.02796622117360433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,6144,0.024345777100986902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,7168,0.02787644333309597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,5120,0.023688889212078516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,5120,0.02212711175282796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,5120,0.02184533410602146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,4096,0.019304000669055514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,4096,0.018944889307022095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,3584,0.017816000514560275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,3584,0.01761511133776771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,3072,0.01553422212600708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,3072,0.015535111228624979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,3072,0.01611822181277805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,4096,0.01850222216712104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,3584,0.0170133329100079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,2560,0.014148443937301636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,2560,0.014360888136757744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,2560,0.014780445231331719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,2048,0.012552888856993781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,2048,0.01257511145538754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,2048,0.013373333546850415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,1536,0.010437332921557956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,1536,0.011847111086050669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,1024,0.009106666677527958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,1024,0.008251555263996124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,1024,0.0090844440791342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,768,0.007078222102589077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,768,0.008620444271299574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,1536,0.011208888557222156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,512,0.005759999983840519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,512,0.007815111014578078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,768,0.008054222497675154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,512,0.006679111056857639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,256,0.00442844463719262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,256,0.007185777856243982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,128,0.0038613333470291565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,128,0.006728888799746831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,64,0.0035946667194366455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,64,0.0069546666410234236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3584,32,0.003607999947335985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3584,32,0.007102222078376346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,65536,0.19704622692532012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,65536,0.2159902254740397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,256,0.005768888940413793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,16384,0.060850666628943555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,65536,0.18841332859463164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3584,128,0.0053048887186580235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,16384,0.05073600014050802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,12288,0.04106311003367106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,16384,0.052988443109724254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,12288,0.04846577843030294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,12288,0.04059377643797133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,10240,0.04248977700869242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,10240,0.035312000248167254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,8192,0.029680887858072918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,8192,0.03219199842876858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,8192,0.02937600016593933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,7168,0.02777155571513706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,7168,0.02716711163520813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,6144,0.02418400016095903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,6144,0.024359999431504145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,10240,0.03557777735922072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,7168,0.027102222045262653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,5120,0.021235555410385132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,5120,0.020647999313142564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,5120,0.021828444467650518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,4096,0.01718844473361969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,4096,0.018798222144444782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,3584,0.01640977793269687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,6144,0.024126221736272175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,3584,0.015268445014953613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,3584,0.01738844480779436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,3072,0.015196444259749519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,3072,0.01366222235891554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,3072,0.016031111280123394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,2560,0.01368799971209632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,4096,0.01790577835506863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,2560,0.014711111783981323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,2048,0.011635555161370171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,2048,0.013182222015327878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,1536,0.010943111446168689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,2560,0.013997332917319404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,2048,0.012611555556456247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,1536,0.009610666996902889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,1536,0.01126488877667321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,1024,0.007708444363541073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,1024,0.009016888837019602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,768,0.006474666711356904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,768,0.008387555678685507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,512,0.0053431110249625305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,1024,0.009317333499590555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,512,0.007814221911960179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,768,0.007560888926188151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,512,0.006430222342411677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,256,0.0042204442951414315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,256,0.00562755556570159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,256,0.007171555525726742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,3072,128,0.005251555393139522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,128,0.0038524443904558816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,128,0.006709333509206772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,64,0.003485333174467087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,64,0.006908444480763541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,3072,32,0.003576888806290097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,3072,32,0.007017778025733099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,65536,0.19607555866241455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,16384,0.0427653325928582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,16384,0.05277333325809903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,65536,0.18480532699161104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,16384,0.05126666691568163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,65536,0.1574808888965183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,12288,0.04270133376121521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,12288,0.0399982233842214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,10240,0.02921600143114726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,12288,0.03369244601991441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,10240,0.03289599882231818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,10240,0.03484444485770331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,8192,0.02459555533197191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,8192,0.027299554811583623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,8192,0.02945244312286377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,7168,0.02234755622016059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,7168,0.024459555745124817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,7168,0.026838221483760413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,6144,0.02012800011369917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,6144,0.02121777832508087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,6144,0.02429955535464817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,5120,0.01830666595035129
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,5120,0.021432888176706102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,4096,0.015230221880806817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,4096,0.014984889162911309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,4096,0.018526222970750596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,3584,0.013258666627936892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,3584,0.017264000243610807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,3072,0.011936888926559024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,3072,0.015903110305468243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,2560,0.012011555333932241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,5120,0.017994667092959087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,2560,0.010651555326249866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,3584,0.01424088908566369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,2560,0.01460000044769711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,2048,0.010716444088353051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,3072,0.013054221868515015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,2048,0.009206222163306342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,2048,0.013029333617952136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,1536,0.007941332956155142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,768,0.00574222207069397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,1536,0.010970667004585266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,1024,0.006590222318967183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,1024,0.008988444175985124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,768,0.006491555521885554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,768,0.008398222426573435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,512,0.005889777921968036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,512,0.0047360000511010485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,512,0.007705777883529663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,256,0.005342222336265776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,128,0.006665777828958299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,1536,0.0094035557574696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,256,0.004008888784382078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,1024,0.007307555940416124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,256,0.007133333219422235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,128,0.0037191112836201987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2560,128,0.004959110998445087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,64,0.0033653332955307434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,64,0.006778666542636023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2560,32,0.0034622223012977173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2560,32,0.0069102222720781965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,65536,0.15962489446004233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,65536,0.14592356152004665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,16384,0.04402755697568258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,16384,0.043290668063693576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,65536,0.18405777878231475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,16384,0.05100533366203308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,12288,0.033634665939542986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,12288,0.03976888789070977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,10240,0.028916445043351915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,10240,0.02922755479812622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,10240,0.03459111187193129
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,12288,0.03359199894799127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,8192,0.029193778832753498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,8192,0.02920711040496826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,8192,0.024350222614076402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,7168,0.025215999947653875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,7168,0.02667288813326094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,6144,0.021919111410776775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,6144,0.024136000209384497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,5120,0.017754667335086398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,5120,0.018590221802393597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,5120,0.021578666236665513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,4096,0.01496444476975335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,4096,0.016170667277442086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,4096,0.018568888306617737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,7168,0.022266666094462078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,3584,0.014044443766276041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,6144,0.019932443896929424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,3584,0.013952889376216464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,3584,0.017287999391555786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,3072,0.012676444318559436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,3072,0.012283555335468717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,3072,0.015855110353893705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,2560,0.009584888815879822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,2560,0.014359111587206522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,2048,0.008445333275530074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,1536,0.010454222559928894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,2048,0.012416000167528788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,1536,0.007855999800893996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,1536,0.007159999675220913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,2560,0.011583999627166323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,1024,0.005543111099137201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,1024,0.008839111361238692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,2048,0.010006222460005019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,768,0.006105777704053455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,1024,0.006678222368160884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,768,0.004933333231343163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,768,0.008301333420806462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,512,0.005603555589914322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,512,0.004400889078776042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,512,0.007651555869314406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,256,0.005165333135260476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,256,0.0038088887102074097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,256,0.007048000064161088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,2048,128,0.00480266660451889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,128,0.003495111233658261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,128,0.006657777975002925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,64,0.0032799999333090256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,64,0.006774222271309958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,2048,32,0.003286222202910317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,2048,32,0.006862222320503659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,65536,0.1333128876156277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,65536,0.11550045013427734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,16384,0.03771555423736572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,16384,0.03463644451565213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,65536,0.18506222301059297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,16384,0.05048000150256687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,12288,0.02944799926545885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,12288,0.027813331948386297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,12288,0.03969066672854953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,10240,0.024871110916137695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,10240,0.03471111257870992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,8192,0.02162222233083513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,8192,0.028853333658642236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,10240,0.026165333059098985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,8192,0.021783111823929682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,7168,0.017829333742459614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,7168,0.02645955483118693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,6144,0.01787911189926995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,6144,0.01588888963063558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,6144,0.023800000548362732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,7168,0.019984000258975558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,5120,0.013924444715181986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,5120,0.021274665991465252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,4096,0.011782222323947482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,4096,0.01852888862291972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,3584,0.011015111373530494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,3584,0.017129777206314933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,3072,0.009743111001120673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,3072,0.015627556376987033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,2560,0.009670221971140968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,5120,0.015951999359660678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,2560,0.01399466726515028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,2048,0.012064889073371887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,4096,0.013558222187889947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,2048,0.008431110945012834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,1536,0.007679111427730984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,1536,0.0069884442620807225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,1536,0.010471111370457543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,3584,0.012696888711717395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,3072,0.011562666959232755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,2560,0.010404444403118556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,2048,0.008813333180215623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,1024,0.006976000136799282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,1024,0.005404444618357553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,1024,0.008832889298597971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,768,0.004804444395833545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,768,0.008229333493444655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,512,0.0041413332025210066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,512,0.007652444144090016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,768,0.006159111029571957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,512,0.005629333357016246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,256,0.003663111064169142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,256,0.007053333024183909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,128,0.00332266671790017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,128,0.006663110935025745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,64,0.00309244439833694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,64,0.006754666566848755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1536,32,0.003162666741344664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,256,0.005085333353943295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1536,32,0.006824888702895906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,65536,0.10776978068881565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,65536,0.07971022526423137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,16384,0.030866665972603693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,16384,0.027015111512607996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,65536,0.18366755379570854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,16384,0.05026222268740336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,12288,0.021451556020312842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,12288,0.03976622223854065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1536,128,0.0047644442982143825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,10240,0.021900445222854614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,10240,0.03442400031619602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,8192,0.018152889278199937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,8192,0.019867555962668527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,8192,0.029005332125557795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,7168,0.015275556180212231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,7168,0.026379555463790894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,6144,0.013609777722093793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,6144,0.02375288969940609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,5120,0.013731555806265937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,12288,0.02422933280467987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,5120,0.012034667034943899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,5120,0.021176000436147053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,4096,0.010416888528399997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,4096,0.018233777748213876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,3584,0.009637333452701569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,3584,0.0170160002178616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,3072,0.008797333472304875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,10240,0.021581333544519212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,7168,0.016785777277416654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,6144,0.015160888433456421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,4096,0.011578666667143503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,3072,0.01514400045077006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,3584,0.010810666614108615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,3072,0.009527111219035255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,2560,0.007996444072988298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,2560,0.013144888811641268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,2048,0.007108444141017065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,2048,0.01147822207874722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,1536,0.005738666488064661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,2560,0.008584889272848764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,2048,0.007522666619883643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,1536,0.010017777482668558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,1024,0.005984888722499211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,1536,0.006877333339717653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,1024,0.004826666580306159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,1024,0.008783111141787635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,768,0.005592888842026393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,768,0.004367111043797599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,768,0.008186666501892937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,512,0.0052728888889153796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,512,0.0038444445365005066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,512,0.007538666327794393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,256,0.0034142221427626083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,256,0.0070053330726093715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,256,0.004831999954250124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,128,0.003160888950030009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,1024,128,0.004575110971927643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,128,0.00656355544924736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,64,0.003015999992688497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,64,0.006711999989218182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,1024,32,0.00309244439833694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,1024,32,0.006714666469229593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,65536,0.1065982249048021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,65536,0.06272177563773261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,16384,0.030421333180533514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,65536,0.18464533487955728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,16384,0.02162666618824005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,16384,0.04990666773584154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,12288,0.02403733299838172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,12288,0.01885066595342424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,12288,0.03975555631849501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,10240,0.016840888394249808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,10240,0.0344782206747267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,8192,0.017716444200939603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,8192,0.01567911108334859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,8192,0.029010666741265193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,7168,0.01644088824590047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,7168,0.014565333724021912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,10240,0.021353777911927965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,7168,0.026259554757012263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,6144,0.013416888813177744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,6144,0.02362666692998674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,5120,0.013562666873137156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,5120,0.012596444123321109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,5120,0.021164443757798936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,4096,0.01131199962562985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,4096,0.011447110937701331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,6144,0.014864888456132678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,4096,0.01809599995613098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,3584,0.011036444041464063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,3584,0.016581333345837064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,3072,0.008612444831265343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,3072,0.014663999279340109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,2560,0.008336000144481659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,2560,0.007672888537247975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,2560,0.012783110969596438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,2048,0.007426666716734569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,2048,0.00665511108107037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,3584,0.010299555957317352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,3072,0.009110222260157267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,2048,0.01124533348613315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,1536,0.005503111001518037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,1536,0.010068444742096795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,1024,0.004642666627963384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,1024,0.008769777913888296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,1536,0.006810666786299811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,768,0.004325333154863781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,768,0.008190222084522247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,512,0.005114666703674528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,512,0.0037884443170494507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,1024,0.005902222047249476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,768,0.005579555614127054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,512,0.00756888919406467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,256,0.0034355556385384668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,256,0.0069875551594628235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,256,0.0047155556579430895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,768,128,0.004510222209824456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,128,0.0031582222630580268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,128,0.006619555668698416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,64,0.002977777686383989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,64,0.006717333363162146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,768,32,0.00304888892504904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,768,32,0.00666222224632899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,65536,0.04446755515204536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,65536,0.1834284464518229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,16384,0.016920000314712524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,65536,0.09682577848434448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,16384,0.050344000260035195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,12288,0.014290667242474027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,12288,0.039628442790773176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,10240,0.014613333675596448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,10240,0.03447644578086005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,8192,0.012982221941153208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,8192,0.028992888000276353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,7168,0.01220266686545478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,7168,0.02645333276854621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,6144,0.011323555476135679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,6144,0.023782221807373896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,5120,0.012328888807031842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,16384,0.028032888968785603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,5120,0.01054222219520145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,5120,0.021193777521451313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,12288,0.022135110365019903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,10240,0.01953066719902886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,8192,0.016574222180578444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,4096,0.00979288915793101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,4096,0.017456889152526855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,3584,0.009343999955389235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,3584,0.008935111264387766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,3584,0.015710221396552194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,3072,0.008046222229798635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,3072,0.01386844449573093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,7168,0.015323556131786771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,6144,0.013971555564138623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,4096,0.01020088874631458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,2560,0.00683822234471639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,3072,0.00849511143234041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,2560,0.012690666649076672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,2048,0.005807111246718301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,2048,0.011298666397730509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,1536,0.006528000036875407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,1536,0.005119110975000593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,2560,0.007749333149856991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,1536,0.010079111489984725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,1024,0.00572622236278322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,1024,0.004359999878538979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,1024,0.008802666432327693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,768,0.005474666754404704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,768,0.004035555654101902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,2048,0.0070293330483966405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,768,0.008240888516108194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,512,0.005011555635266834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,512,0.003674666500753827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,512,0.00756444450881746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,256,0.0033173333439562055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,256,0.006959111326270633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,256,0.004680888934267892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,128,0.003044444446762403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,128,0.006585777633719974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,512,128,0.004417777889304691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,64,0.0028328889360030494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,64,0.00673333348499404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,512,32,0.0029297777348094513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,512,32,0.006706666615274217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,65536,0.09298133187823826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,65536,0.0313368903266059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,16384,0.026310222016440496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,16384,0.013060444758998023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,65536,0.18591199980841744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,16384,0.04999377661281162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,12288,0.02120888895458645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,12288,0.01127199994193183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,12288,0.03964799973699782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,10240,0.010468444062603844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,10240,0.03431288732422723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,8192,0.015672889020707872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,8192,0.009523555636405945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,8192,0.028856890069113836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,7168,0.014515555567211576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,7168,0.009195555415418413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,7168,0.02612977723280589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,10240,0.018597332967652213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,6144,0.008753777378135258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,6144,0.023564444647894964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,5120,0.011114666859308878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,5120,0.009658666948477427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,5120,0.020296888218985665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,4096,0.009570666485362584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,4096,0.008363555702898238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,4096,0.01685599982738495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,3584,0.008926221893893348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,3584,0.007894222107198503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,3584,0.015237333046065437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,6144,0.012695111334323883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,3072,0.007289778027269576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,3072,0.013771555489963956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,2560,0.007502222226725684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,2560,0.0063680000603199005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,3072,0.008182221816645728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,2560,0.012512000070677863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,2048,0.006808888994985157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,2048,0.00566133318675889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,2048,0.011151110960377587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,1536,0.004975999808973736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,1536,0.010000000397364298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,1024,0.00555377784702513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,1024,0.004262222184075249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,1024,0.008646222452322641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,768,0.005250666704442766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,768,0.0038853333228164246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,768,0.008079110748238033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,512,0.004908444566859139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,512,0.003528000166018804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,1536,0.0063075555695427795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,512,0.007459555235173967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,256,0.004560889055331548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,256,0.003136000078585413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,256,0.006839111033413146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,256,128,0.004269333349333869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,128,0.0028844444702068963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,128,0.006496888895829518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,64,0.0027431110954946945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,64,0.006614222294754452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,256,32,0.0027875554644399216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,256,32,0.00654666663871871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,65536,0.026943999860021803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,65536,0.18604444132910836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,16384,0.026240888569090102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,16384,0.010456889039940305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,16384,0.05020088950792948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,12288,0.021007999777793884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,65536,0.09232622385025024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,12288,0.010381333529949188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,12288,0.03935733437538147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,10240,0.009752000371615091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,10240,0.03412444392840067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,8192,0.015628443823920358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,8192,0.008915555146005418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,10240,0.0184888889392217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,8192,0.028592887851927016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,7168,0.008450667063395182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,7168,0.026233777403831482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,6144,0.01218844453493754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,6144,0.007458666960398356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,6144,0.022993778189023335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,5120,0.007491555478837755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,5120,0.019578667150603402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,4096,0.006831999868154526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,4096,0.016342222690582275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,7168,0.013734222286277346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,3584,0.008850666383902231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,3584,0.00720711135201984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,3584,0.014967110421922473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,5120,0.011001778145631155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,3072,0.008061333662933773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,4096,0.009458666874302758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,3072,0.0069466663731469055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,3072,0.013786666923099093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,2560,0.007367110914654202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,2560,0.006280888699822956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,2560,0.012440889245933957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,2048,0.005586666779385672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,2048,0.01111288865407308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,1536,0.006263999889294307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,1536,0.004879111217127906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,1536,0.009926222264766693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,1024,0.005438222239414851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,1024,0.0041448887851503165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,1024,0.008626666333940294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,768,0.005164444446563721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,768,0.003834666477309333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,768,0.008040000167157915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,512,0.0034737777378824023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,512,0.00739644467830658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,256,0.004433777597215441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,256,0.003118222165438864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,256,0.0068862222962909276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,2048,0.006627555522653792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,128,0.002858666703104973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,128,0.006471111128727595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,64,0.0027520000520679685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,64,0.006597333484225803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,128,32,0.00281866660548581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,128,32,0.006585777633719974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,65536,0.023825777901543513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,65536,0.18518843915727404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,512,0.0047395556337303585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,16384,0.00941955546538035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,384,128,128,0.004269333349333869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,16384,0.050082667006386646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,12288,0.008589333130253686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,12288,0.03939466675122579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,10240,0.008793777889675563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,10240,0.0338986681567298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,8192,0.007736889024575551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,8192,0.02853511108292474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,7168,0.007373332977294922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,6144,0.007179555793603261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,7168,0.026064889298544988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,6144,0.022835556003782485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,5120,0.007355555891990662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,5120,0.019318222999572754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,4096,0.006695999867386288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,4096,0.016209777858522203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,3584,0.007041778001520369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,3584,0.015059555570284525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,3072,0.006807111203670502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,3072,0.013741333451535968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,2560,0.006142222219043308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,2560,0.012478222449620565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,2048,0.0054560001525614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,768,0.008066666622956594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,2048,0.011113777756690979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,1536,0.0047662220895290375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,1536,0.009896888501114314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,1024,0.0041039999988343985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,1024,0.008609777523411645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,768,0.003738666574160258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,512,0.003379555626047982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,512,0.00738933351304796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,256,0.003079111170437601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,256,0.006784889019197888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,128,0.0028168888141711554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,128,0.0063999998900625445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,64,0.00267999991774559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,64,0.006552888701359431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,64,32,0.0027155555370781156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,64,32,0.006511999915043513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,65536,0.022092445029152766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,65536,0.18599732716878256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,16384,0.0099653336736891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,16384,0.04993244343333774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,12288,0.008769777913888296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,12288,0.03907466597027249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,10240,0.008482666479216682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,10240,0.033823112646738686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,8192,0.007481777833567725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,8192,0.028540445698632136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,7168,0.007275555696752336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,7168,0.02596355477968852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,6144,0.006598222172922558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,6144,0.022152889106008742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,5120,0.007076444725195567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,5120,0.01900622248649597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,4096,0.006585777633719974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,4096,0.01622577839427524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,3584,0.006978666616810693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,3584,0.015047111445003085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,3072,0.006631999793979857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,3072,0.013659555878904132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,2560,0.006166222194830577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,2560,0.01237155579858356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,2048,0.005435555759403441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,2048,0.011090666883521609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,1536,0.004767111192146937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,1536,0.009909333454238044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,1024,0.00407288885778851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,1024,0.008566222257084316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,768,0.003740444365474913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,768,0.008040889269775814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,512,0.003383111208677292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,512,0.007326222128338284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,256,0.0030533334033356774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,256,0.006826666494210561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,128,0.0028239999794297745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,128,0.00638133328821924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,64,0.0026791110220882627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,64,0.0064995553758409284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,384,32,32,0.0026808888134029177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,384,32,32,0.0064239998658498125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,16384,0.8569075796339246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,16384,0.43776623407999676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,12288,0.5882826911078559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,10240,0.4871190918816461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,12288,0.3246862093607585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,10240,0.27308532926771373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,8192,0.2197706699371338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,12288,0.54349242316352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,8192,0.39406845304701066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,10240,0.44399023056030273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,16384,0.8111466831631131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,7168,0.3522924317253961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,8192,0.3520195484161377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,6144,0.29490576850043404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,7168,0.3158968819512261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,6144,0.26209243138631183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,5120,0.2529013421800402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,5120,0.20574043856726754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,4096,0.20350844330257842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,4096,0.154874669180976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,7168,0.19252888361612955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,6144,0.16982221603393555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,5120,0.14184089501698813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,3584,0.17508089542388916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,3584,0.1026266680823432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,4096,0.11292000611623128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,3584,0.14060978094736734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,3072,0.15689333279927573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,3072,0.09170844157536824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,3072,0.12851911120944554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,2560,0.07645066579182942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,2560,0.1275048918194241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,2560,0.10178222258885701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,2048,0.062237335575951465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,2048,0.10651555326249863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,1536,0.050296889411078564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,2048,0.08393422100279067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,1536,0.08255110846625434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,1536,0.06558222240871854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,1024,0.03728444377581278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,1024,0.05740355783038669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,1024,0.051009777519438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,768,0.031143112315071955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,768,0.04539377821816338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,768,0.04599822229809231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,512,0.02607111136118571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,512,0.03370399938689338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,512,0.04127466678619385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,256,0.020042666130595736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,256,0.02387822171052297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,256,0.038740442858801946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,128,0.01815022197034624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,65536,128,0.016735111673672993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,128,0.03924977779388428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,64,0.014797333214018079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,64,0.04054133428467645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,65536,32,0.015286222100257874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,65536,32,0.04203022188610501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,16384,0.1957484483718872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,12288,0.14538844426472983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,12288,0.08712889088524713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,16384,0.19402400652567545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,65536,0.7885244687398275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,12288,0.11893599563174778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,10240,0.07154933611551921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,10240,0.12459288703070746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,65536,0.8633804321289062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,10240,0.09927111201816136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,8192,0.09952711396747166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,8192,0.06207378043068779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,8192,0.07988266812430488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,7168,0.052426666021347046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,7168,0.09034399853812324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,16384,0.114574220445421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,65536,0.43775733311971027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,7168,0.06779999865425958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,6144,0.04546133346027798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,6144,0.07815111345714994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,6144,0.05784889062245687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,5120,0.03924177752600776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,5120,0.06371644470426771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,5120,0.05015822251637777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,4096,0.03398311138153076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,4096,0.0505582226647271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,4096,0.0403039985232883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,3584,0.030135111676322088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,3584,0.04460710949367947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,3584,0.03639288743336996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,3072,0.026642666922675237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,3072,0.03964266512129042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,3072,0.032474666833877563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,2560,0.02401422295305464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,2560,0.03320088982582092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,2560,0.028875556257035997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,2048,0.020575111111005146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,2048,0.027519999278916255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,2048,0.025412445267041523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,1536,0.02172800070709652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,1536,0.02089333368672265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,1536,0.01806400054030948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,1024,0.015018666783968607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,1024,0.015843555331230164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,1024,0.017699556218253244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,768,0.013212444053755866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,768,0.01607644392384423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,512,0.010695999695195092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,512,0.014906666345066495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,256,0.010528888967302112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,768,0.013348444468445249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,512,0.011897777517636618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,256,0.00811644477976693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,256,0.012278222375445895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,16384,128,0.009078222016493479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,128,0.005950222412745158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,128,0.010863111250930362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,64,0.005193777796294955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,64,0.011424889167149862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,16384,32,0.00563466673096021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,16384,32,0.013574221895800697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,16384,0.08347111278110081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,16384,0.1558497746785482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,16384,0.14117511113484701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,12288,0.11327821678585476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,65536,0.6159831153021919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,65536,0.6324773364596897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,12288,0.09072444174024795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,10240,0.09716977675755818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,12288,0.06283199787139893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,65536,0.32548533545600045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,10240,0.07411466704474555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,8192,0.07878844605551825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,8192,0.060473779837290444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,7168,0.0671644475724962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,7168,0.05225155419773526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,6144,0.03451199995146857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,6144,0.05961689021852282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,6144,0.0443484452035692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,10240,0.05490755703714159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,8192,0.0444160004456838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,5120,0.02998222245110406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,5120,0.04948888884650337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,7168,0.039905776580174766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,5120,0.03897066579924689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,4096,0.02572800053490533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,4096,0.03947910997602675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,4096,0.03241066800223456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,3584,0.03451022174623277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,3584,0.02921600143114726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,3072,0.030037333567937214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,3584,0.02348088887002733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,3072,0.026625777284304302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,3072,0.02129688858985901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,2560,0.026339555780092876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,2560,0.023797333240509033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,2048,0.016948444975747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,2048,0.021242666575643752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,2048,0.021359110871950786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,2560,0.01919911139541202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,1536,0.016957332690556843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,1536,0.017785777648289997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,1024,0.012503110700183444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,1024,0.012701333396964602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,1536,0.014942222171359591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,1024,0.015136889285511441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,768,0.010834666589895884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,768,0.013946667313575745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,512,0.010423111418883005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,512,0.008791999684439765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,512,0.012554667062229581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,256,0.009249777429633671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,256,0.0064080001579390625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,256,0.010220444036854638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,768,0.011415999796655444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,12288,128,0.007857778006129796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,128,0.005028444445795483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,128,0.009557333257463243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,64,0.004464888738261329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,64,0.00997866690158844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,12288,32,0.0048142220411035745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,12288,32,0.010996444357766045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,65536,0.28229422039455837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,65536,0.5099715656704372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,65536,0.5275057686699761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,16384,0.12839644485049778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,16384,0.07102933194902208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,16384,0.1177519957224528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,12288,0.0967937774128384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,12288,0.07276888688405354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,10240,0.04941866795221964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,12288,0.054883556233512036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,10240,0.08310400115119086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,10240,0.06104711029264662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,8192,0.06653955247667101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,8192,0.05052622159322103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,7168,0.03525688913133409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,7168,0.05793155564202202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,7168,0.04467022087838915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,8192,0.039057777987586126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,6144,0.031048887305789526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,6144,0.04982399940490723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,6144,0.03846577803293864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,5120,0.026927999324268762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,5120,0.042103999190860324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,5120,0.0336151123046875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,4096,0.033596442805396184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,4096,0.02277333372169071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,4096,0.02827377782927619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,3584,0.02107200026512146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,3584,0.03053599927160475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,3584,0.025847110483381484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,3072,0.018974221414989896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,3072,0.02658933401107788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,2048,0.01864266726705763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,3072,0.023223999473783705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,2560,0.023079999619060095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,2560,0.020983111527231004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,2048,0.01534933348496755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,2048,0.01908177799648709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,2560,0.017437333861986797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,1536,0.013830222189426422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,1536,0.015208888385030957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,1536,0.01566133399804433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,1024,0.011598221957683563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,1024,0.0119991112086508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,1024,0.013608888619475894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,768,0.010431110858917236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,768,0.010270222193664974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,768,0.012643555800120035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,512,0.00868088917599784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,512,0.011221333510345884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,512,0.009469333622190688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,256,0.006241777704821692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,256,0.008153777983453538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,256,0.009402666654851701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,10240,128,0.006768888897365994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,128,0.004948444250557157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,128,0.00886933339966668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,64,0.004417777889304691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,64,0.009143999881214565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,10240,32,0.004635555462704765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,10240,32,0.010011555420027839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,65536,0.3932986789279514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,16384,0.10283733076519436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,16384,0.09217777517106797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,12288,0.07847288582060072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,12288,0.05857866340213352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,65536,0.42417155371771914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,10240,0.039679110050201416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,10240,0.06407999992370605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,10240,0.049807111422220864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,8192,0.05307111144065857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,8192,0.04118577639261881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,65536,0.2334986792670356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,7168,0.029467556211683486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,7168,0.046345776981777616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,7168,0.03650666607750787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,6144,0.025951999757024977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,16384,0.06176444556978014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,12288,0.04671288861168755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,8192,0.0328106681505839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,6144,0.04046400056944953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,6144,0.032235556178622775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,5120,0.022537777821222942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,5120,0.033835556772020124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,5120,0.028258668051825628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,4096,0.019560888409614563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,4096,0.027553778555658128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,4096,0.024059555596775476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,3584,0.023924445112546284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,3584,0.02197422252760993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,3072,0.02128000060717265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,3072,0.019813333948453266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,2560,0.0181688881582684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,2560,0.01795022189617157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,3584,0.018258665998776753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,3072,0.016571554872724745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,2560,0.015096000499195524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,2048,0.015185778339703878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,2048,0.016044444508022733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,1536,0.012083555261294046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,1536,0.012167111039161682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,2048,0.013460444079505073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,1536,0.01401600076092614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,1024,0.009559999737474654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,1024,0.01220533334546619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,768,0.008392888638708327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,768,0.011016888750924004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,512,0.008719999757077958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,512,0.0069475554757648045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,512,0.009197333620654212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,256,0.00666133314371109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,256,0.005051555732885997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,256,0.008552889029184977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,128,0.0058266665372583605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,1024,0.010210666391584609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,128,0.00434755575325754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,8192,768,0.009653333160612319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,128,0.008058666355080074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,64,0.004602666530344221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,64,0.008656000097592672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,8192,32,0.0046897778908411665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,8192,32,0.008969777988062965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,65536,0.3634035587310791
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,65536,0.2395111189948188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,65536,0.379347562789917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,16384,0.09129066599739923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,16384,0.062189333968692355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,16384,0.08571377727720474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,12288,0.06937155458662245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,12288,0.05678133169809977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,10240,0.03992266787423028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,10240,0.059157331784566246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,10240,0.04924355612860786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,12288,0.04621600111325582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,8192,0.045752889580196805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,8192,0.040130668216281466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,7168,0.0296106669637892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,7168,0.04105244411362542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,7168,0.0358168880144755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,6144,0.025632888078689575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,6144,0.03467822074890137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,6144,0.03204800022972955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,5120,0.02259911100069682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,5120,0.029719998439153034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,5120,0.02808800008561876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,4096,0.019493333167499967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,8192,0.03310488992267185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,4096,0.023927999867333308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,4096,0.023909333679411147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,3584,0.018254222141371835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,3584,0.021711111068725586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,3584,0.022056889202859666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,3072,0.019183999962276883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,3072,0.01964000033007728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,2560,0.016176000237464905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,2560,0.01794933279355367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,2048,0.013209777573744455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,2048,0.013559111290507846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,2048,0.016024889217482675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,1536,0.011209777659840055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,1536,0.013744889034165276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,1024,0.010618666807810465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,1024,0.008776889079146916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,1024,0.012113778127564324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,3072,0.016353777713245816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,2560,0.015063110325071545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,768,0.007679111427730984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,768,0.010586666564146677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,512,0.008616000413894653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,512,0.0064479998416370815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,1536,0.012053333222866058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,512,0.009259555902745988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,256,0.004819555415047539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,256,0.00852888905339771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,128,0.006079999936951532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,128,0.004120888809363048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,128,0.008124444219801161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,64,0.0038613333470291565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,768,0.009472889204819998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,7168,256,0.00647555540005366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,64,0.008263111114501953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,7168,32,0.003958222352796131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,7168,32,0.009017777939637503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,16384,0.08255644639333089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,65536,0.3158666557735867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,65536,0.21244088808695474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,16384,0.06516533427768283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,65536,0.3184097872840034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,12288,0.05968088573879666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,12288,0.04516177707248264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,10240,0.040170666244294904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,16384,0.05561955769856771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,10240,0.05148799882994758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,12288,0.043303999635908336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,10240,0.03890577620930142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,8192,0.040545778142081365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,8192,0.03316977620124817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,7168,0.03520977828237746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,7168,0.029322667254341975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,6144,0.02462844385041131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,8192,0.031234665049446955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,7168,0.028198222319285076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,6144,0.030902223454581365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,6144,0.02648799949222141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,5120,0.025888888372315302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,5120,0.02346844474474589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,4096,0.01871288816134135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,4096,0.02160266704029507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,4096,0.019895111521085102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,3584,0.019533332851197984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,3584,0.018403554956118267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,3072,0.01581333412064446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,3072,0.01683022247420417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,5120,0.021735999319288466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,3072,0.016728000508414376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,3584,0.01740622189309862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,2560,0.01462311049302419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,2560,0.014474666780895658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,2560,0.015493333339691162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,2048,0.012973333398501078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,2048,0.0120666664507654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,2048,0.013926222920417786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,1536,0.010264889233642155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,1536,0.012238221863905588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,1024,0.008114666574531132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,1536,0.011559111376603445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,1024,0.010339555641015371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,768,0.00870488915178511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,768,0.007116444408893585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,768,0.009111999637550777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,512,0.0072266666425599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,512,0.00572622236278322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,1024,0.009910222556855943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,512,0.008276444342401292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,256,0.004471999903519948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,256,0.007712889048788283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,32,0.0038213332494099936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,128,0.0053102220926019884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,128,0.004027555386225383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,6144,256,0.005769777629110549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,128,0.007443555527263218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,6144,64,0.0036764442920684814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,64,0.007640889121426477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,6144,32,0.00815822184085846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,65536,0.18490755558013916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,65536,0.25809065500895184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,65536,0.2987884415520562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,16384,0.05106133222579956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,16384,0.07226222091250949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,16384,0.06318666537602742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,12288,0.052503999736573964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,12288,0.04509511258867052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,10240,0.03575377662976583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,12288,0.039649777942233615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,10240,0.04404177930619982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,10240,0.03745511174201965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,8192,0.0384880006313324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,8192,0.0313937763373057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,7168,0.025615110993385315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,7168,0.0314631097846561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,7168,0.029438220792346533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,6144,0.022522666388087805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,6144,0.02734577821360694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,6144,0.025952888859642878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,8192,0.028756442997190688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,5120,0.020080000162124634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,5120,0.02362666692998674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,5120,0.023342221975326538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,4096,0.021394666698243883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,4096,0.019661333825853135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,3584,0.018603555030292936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,3584,0.01815733313560486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,3072,0.015443555182880826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,4096,0.01734844512409634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,3584,0.016269332832760282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,3072,0.01681866745154063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,2560,0.013631111217869652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,2560,0.013747555514176687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,2560,0.015268445014953613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,2048,0.011328889264000786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,3072,0.014710222681363424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,2048,0.013861333330472311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,2048,0.012079111403889127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,1536,0.00998044427898195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,1536,0.012127111355463663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,1024,0.009019555317031013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,1024,0.00739733295308219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,1024,0.0098213329911232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,1536,0.01074399964676963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,768,0.006893333461549547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,768,0.008837333156002892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,768,0.007746666669845581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,512,0.006258666515350342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,512,0.005113777601056629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,512,0.008080888953473832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,256,0.004370666626426908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,256,0.007639110916190677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,128,0.0038862224254343244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,128,0.007335999773608313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,64,0.0035751110149754416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,64,0.007520000139872233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,5120,32,0.0037679999238914917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,5120,32,0.007857778006129796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,65536,0.2003413306342231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,65536,0.21051288975609672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,256,0.005508444375462002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,5120,128,0.005097777893145879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,65536,0.1526986625459459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,16384,0.05756533145904541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,16384,0.0409457782904307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,16384,0.044234666559431285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,12288,0.032144887579811945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,12288,0.04374577932887607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,12288,0.03272799981964959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,10240,0.027990221977233887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,10240,0.038020445240868464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,10240,0.028920888900756836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,8192,0.030798223283555772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,8192,0.02463022205564711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,7168,0.021929777330822412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,7168,0.02658133374320136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,7168,0.02221333318286472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,6144,0.02402399977048238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,8192,0.024002666274706524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,6144,0.020296888218985665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,5120,0.017118222183651395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,6144,0.019168888529141743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,5120,0.021346666746669348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,5120,0.01813155578242408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,4096,0.014911111858155994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,4096,0.01624000072479248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,4096,0.015936889582210116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,3584,0.014030221435758801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,3584,0.014696000350846184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,3584,0.014674666855070325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,3072,0.013897778259383308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,3072,0.013637333280510373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,3072,0.012912888493802814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,2560,0.011952000359694162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,2560,0.012362666428089142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,2560,0.012646222280131446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,2048,0.010271111296282874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,2048,0.011617778076065911
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,1536,0.00983199973901113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,2048,0.010748444332016839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,1536,0.008817777865462834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,1536,0.010015111002657149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,1024,0.007899555895063613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,1024,0.0068897778789202375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,1024,0.008017777568764156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,768,0.006800000038411882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,768,0.006039999839332368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,768,0.0076044441925154785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,512,0.004974222017659081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,256,0.005349333501524395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,512,0.007179555793603261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,256,0.004056000047259861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,256,0.006800888727108638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,128,0.004923555586073133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,128,0.003630222131808599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,128,0.006507555643717448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,64,0.003581333491537306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,64,0.006746666712893381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,4096,32,0.0036311112344264984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,4096,32,0.007298666569921706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,4096,512,0.005790222022268508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,65536,0.1799022224214342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,16384,0.050831112596723765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,16384,0.042669332689709134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,65536,0.20565511120690239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,12288,0.04029777646064758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,12288,0.03231466809908549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,10240,0.034727109803093806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,65536,0.15690933333502874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,10240,0.02964977754486932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,8192,0.024958221448792353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,8192,0.02978044417169359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,8192,0.02389066749148899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,7168,0.024493333366182115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,7168,0.022456000248591106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,16384,0.04302311274740431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,6144,0.01962222158908844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,6144,0.021361778179804485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,12288,0.03351022137535943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,10240,0.02901244494650099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,6144,0.020170667105250888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,5120,0.017536888519922893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,5120,0.018197332819302876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,5120,0.0181022220187717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,7168,0.02279111080699497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,4096,0.015081778168678284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,4096,0.01589777734544542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,3584,0.013390222357379066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,3584,0.014726221561431885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,3072,0.01204355557759603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,3072,0.013718222578366598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,2560,0.011334222224023608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,2560,0.012705778082211813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,4096,0.015239111251301236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,2048,0.009719111025333405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,3584,0.014270222849316068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,2048,0.011702222128709158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,1536,0.00977244476477305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,1536,0.007658667034573025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,1536,0.00958577791849772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,3072,0.013008000122176277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,2560,0.012155555188655853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,2048,0.01109688894616233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,1024,0.006559999866618051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,1024,0.008270222279760573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,768,0.006606222026877933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,768,0.005740444279379315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,768,0.007651555869314406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,512,0.004479111068778568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,512,0.007224889265166388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,256,0.0038373333712418876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,256,0.007257777783605788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,128,0.00498844434817632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,128,0.003571555432346132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,128,0.006569777925809224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,64,0.0032879999942249725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,64,0.006755555669466655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3584,32,0.0033280000918441345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3584,32,0.007096000015735626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,65536,0.15899644957648382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,16384,0.043041778935326465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,65536,0.1515208880106608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,65536,0.197343111038208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,1024,0.00812000036239624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,16384,0.04455733299255371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,16384,0.04010666741265191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,512,0.005922666854328579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,12288,0.03595111105177138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3584,256,0.0053688887920644544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,12288,0.03223377797338698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,12288,0.03282311227586534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,10240,0.031303998496797346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,10240,0.028068443139394123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,8192,0.0240346673462126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,8192,0.027434666951497395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,8192,0.024142222272025213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,7168,0.021976888179779053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,7168,0.02167466613981459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,10240,0.028231110837724473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,7168,0.022340445054901972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,6144,0.019330667124854196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,6144,0.019370666808552213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,6144,0.019888889458444383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,5120,0.017143999536832173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,3584,0.01275111072593265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,5120,0.01647200021478865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,5120,0.017802667286660936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,4096,0.014885332849290637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,4096,0.015124445160230001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,4096,0.01556799974706438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,3584,0.013815999858909182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,3584,0.014378666877746582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,3072,0.011014222270912595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,3072,0.013429332938459186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,2560,0.011749332977665795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,2560,0.010126222338941362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,2560,0.01258933295806249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,2048,0.008584000170230865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,2048,0.011305777562989129
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,1536,0.00740444411834081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,1536,0.00942755573325687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,1024,0.006610666712125142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,1024,0.006044444524579578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,1024,0.008103110724025303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,768,0.004944889081848992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,768,0.007677333222495184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,512,0.004257777912749184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,3072,0.012700444294346703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,2048,0.010288889209429422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,512,0.007259555160999298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,1536,0.008617777791288164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,256,0.00370399985048506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,768,0.005992888990375731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,256,0.006787555499209299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,512,0.005442666510740916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,128,0.004755555755562252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,128,0.0034186666210492453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,128,0.006504000061088138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,64,0.0032337777730491427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,64,0.006680000159475539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,3072,256,0.005137777576843898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,3072,32,0.003338666632771492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,3072,32,0.006945778098371293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,65536,0.12744088967641196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,65536,0.13357778390248617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,65536,0.17730132738749185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,16384,0.035895999934938215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,16384,0.04036000039842393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,16384,0.039769778649012245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,12288,0.02809777855873108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,12288,0.031792001591788396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,12288,0.031750222047170006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,10240,0.028038223584493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,10240,0.027961777316199407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,8192,0.020788444413079154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,8192,0.026368889543745253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,8192,0.023629332582155865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,7168,0.022436444958051045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,7168,0.02160088883505927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,6144,0.01679999960793389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,6144,0.016493333710564505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,10240,0.02480977773666382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,6144,0.019744000501102872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,7168,0.019104889697498746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,5120,0.015069334043396844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,5120,0.014239999983045789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,5120,0.017816889617178176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,4096,0.012126222252845764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,4096,0.015473778049151102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,3584,0.012467555701732635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,3584,0.011086222198274402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,4096,0.01330044451687071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,3584,0.014454222387737699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,3072,0.011384000380833944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,3072,0.010115555591053432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,3072,0.013382222089502545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,2560,0.009014222357008193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,2560,0.012235555383894177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,2048,0.00924177798959944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,2048,0.007704888780911763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,2048,0.010695999695195092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,2560,0.010796444283591377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,1536,0.006787555499209299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,1536,0.008824889030721452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,1024,0.006666666517655055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,1536,0.007723555796676212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,1024,0.005401777724424998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,1024,0.008062221937709385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,768,0.006103111224042044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,768,0.0046968890560997855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,768,0.007576888634098901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,512,0.005558222118351195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,512,0.00408266650305854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,512,0.007091555330488417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,256,0.005069333232111401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,256,0.00370133337047365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,256,0.006725333217117522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,128,0.003338666632771492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,128,0.006495111104514863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,64,0.0031271111220121384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,64,0.006691555596060223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2560,128,0.004772444566090901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2560,32,0.0032106666929192017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2560,32,0.006813333266311222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,65536,0.10393422179751927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,65536,0.11016177468829685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,16384,0.029543999168607924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,65536,0.10734577973683675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,16384,0.03163289030392965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,16384,0.02656800051530202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,12288,0.023394667439990576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,12288,0.026687999566396076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,12288,0.021425777011447485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,10240,0.024174221687846716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,10240,0.019381332728597853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,8192,0.017808889349301655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,8192,0.022504000200165644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,8192,0.016751110553741455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,7168,0.016488000750541687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,7168,0.019290667441156175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,10240,0.020485333270496793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,7168,0.015631111131774057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,6144,0.014394667413499622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,6144,0.014268444644080268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,6144,0.014561777313550314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,5120,0.013110222087966071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,5120,0.013061333033773633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,5120,0.013406222065289816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,4096,0.012068444656001197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,4096,0.011884444289737277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,3584,0.0107360002067354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,3584,0.011295110815101199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,3072,0.010235555469989777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,3072,0.008861333131790161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,3072,0.010654222634103564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,4096,0.01166311071978675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,2560,0.009076444639099969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,2560,0.008463111188676622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,2560,0.009332444104883406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,3584,0.01109333336353302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,2048,0.007709333466158972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,2048,0.007259555160999298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,2048,0.008092444803979661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,1536,0.00675733346078131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,1536,0.0063457778758472866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,1536,0.007376000285148621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,1024,0.004879999905824661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,1024,0.006862222320503659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,768,0.005704000178310607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,768,0.004567111117972268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,768,0.006688000013430913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,512,0.0052044445441828836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,512,0.004074666649103165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,512,0.006331555545330048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,256,0.00360355567600992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,256,0.005984888722499211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,1024,0.005961777849329843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,128,0.003297777846455574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,32,0.006198222024573221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,128,0.005758222192525864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,64,0.0031235555393828284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,2048,64,0.006028444402747684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,2048,32,0.0032799999333090256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,65536,0.08225333028369479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,256,0.004918222212129169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,65536,0.08753955364227295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,16384,0.028873778051800195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,16384,0.02610844373703003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,16384,0.025383111503389146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,12288,0.022727999422285292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,2048,128,0.004595555365085602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,12288,0.021348444951905146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,12288,0.020735111501481798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,65536,0.09936622116300796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,10240,0.020018666982650757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,10240,0.020592000749376085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,10240,0.01870577699608273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,8192,0.017269333203633625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,8192,0.019482667247454327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,8192,0.016061334146393668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,7168,0.01648000048266517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,7168,0.015122666954994202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,6144,0.01402844488620758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,6144,0.013582222163677216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,6144,0.014161777165200977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,5120,0.012647999657524956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,5120,0.011901333100265928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,7168,0.016128000285890367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,5120,0.013160000244776407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,4096,0.010470222267839642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,4096,0.011556444068749746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,3584,0.010231999887360467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,3584,0.009699555734793345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,3584,0.011039111349317761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,3072,0.009331555830107795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,3072,0.008848889006508721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,4096,0.011197333534558615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,3072,0.009761778016885122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,2560,0.008492444124486711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,2560,0.008040889269775814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,2560,0.009015111459626092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,2048,0.007191999918884701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,2048,0.007669332954618666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,1536,0.006240889016124938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,1536,0.007248889240953658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,1024,0.00481244424978892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,1024,0.006745777610275481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,768,0.004377777791685528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,768,0.006601777755551868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,512,0.0038782221575578055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,512,0.006321777900060018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,256,0.0034426665968365143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,256,0.005991111199061076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,128,0.004579555657174853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,128,0.0032053333189752367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,128,0.005778666585683823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,64,0.003063111048605707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,64,0.005972444597217772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1536,32,0.0030586665703190696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1536,32,0.005967111223273807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,65536,0.05638844437069363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,65536,0.08947288990020752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,65536,0.07857244544559054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,16384,0.02639644510216183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,16384,0.021511110994550917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,16384,0.02425066630045573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,12288,0.017062221964200337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,12288,0.020058666666348774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,2048,0.007231999602582719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,10240,0.018422222799725003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,10240,0.015291555060280694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,10240,0.018032888571421306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,8192,0.015951999359660678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,8192,0.014882667197121514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,1536,0.006676444576846228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,8192,0.01564088960488637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,7168,0.012872888810104795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,7168,0.014696000350846184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,1024,0.005863111052248213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,6144,0.012977778083748288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,6144,0.011922666596041786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,6144,0.013615111509958902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,5120,0.011762667033407422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,5120,0.011047999891969891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,5120,0.012746666868527731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,4096,0.010216889282067617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,4096,0.009949333137936061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,4096,0.01069866700304879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,768,0.005537777725193236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,512,0.005160888863934411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,3584,0.00907733374171787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1536,256,0.004791999856630961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,12288,0.020920889245139226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,7168,0.014820444915029736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,3584,0.010092444717884064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,3072,0.008178667061858708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,3072,0.009069333473841349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,2560,0.007647111184067196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,3584,0.00907733374171787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,3072,0.008157333566082848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,2560,0.007738666401969061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,2560,0.008197333249780867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,2048,0.006789333290523953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,2048,0.007608888877762689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,1536,0.005676444619894028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,1536,0.007126222054163615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,1024,0.005718222094906702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,1024,0.004765333400832282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,1024,0.006654222392373615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,768,0.004335111214054955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,2048,0.0068604445291890045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,1536,0.00639111093348927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,768,0.006424888968467712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,768,0.005377777748637729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,512,0.003768889026509391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,512,0.006134222365087933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,512,0.004990222139490975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,256,0.0047146665553251905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,256,0.0033324443631701996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,256,0.005842666659090254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,1024,128,0.004429333325889376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,128,0.003080888961752256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,128,0.005686222265164058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,64,0.0029351111087534162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,64,0.005697777701748743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,1024,32,0.002921777880854077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,1024,32,0.005787555542257097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,65536,0.08848177724414402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,65536,0.046127110719680786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,65536,0.07840800285339355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,16384,0.01715555621518029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,16384,0.024259555670950148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,12288,0.020506666766272653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,12288,0.014872888724009195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,12288,0.02001066671477424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,10240,0.018269333574506972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,10240,0.012853333519564735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,10240,0.017906667457686532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,8192,0.013203555511103736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,8192,0.01553600033124288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,7168,0.014706666270891825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,16384,0.026015111141734656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,7168,0.01236355553070704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,7168,0.014663111832406787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,8192,0.015860444969601102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,6144,0.012730666332774691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,6144,0.011344888971911537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,6144,0.013640888863139682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,5120,0.010718222293588849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,5120,0.012607110871209038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,4096,0.009477333062224919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,4096,0.009999111294746399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,4096,0.010289777484205034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,3584,0.008997333546479544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,5120,0.011156444748242697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,3584,0.008794666992293464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,3584,0.009560888840092553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,3072,0.008081778056091731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,3072,0.008066666622956594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,3072,0.008830221990744272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,2560,0.007540444533030192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,2560,0.0069679998689227635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,2560,0.00815555536084705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,2048,0.006739555547634761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,2048,0.005966222120655908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,2048,0.007553777760929531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,1536,0.005150222116046482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,1536,0.007078222102589077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,1024,0.005616888817813661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,1024,0.004395555704832077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,1024,0.006617777877383762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,768,0.005285333428117964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,768,0.004030222280157937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,768,0.00638755535085996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,1536,0.006272000157170826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,128,0.004368888835112254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,512,0.004980444494220945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,512,0.0036684444381131064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,512,0.006120000034570694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,768,256,0.004605333424276776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,256,0.0032808888289663526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,256,0.0058417779703934984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,128,0.00300177786913183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,128,0.005659555395444234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,64,0.002862222285734283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,64,0.005731555736727185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,16384,0.014712888333532544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,768,32,0.0029013332807355454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,768,32,0.005744888964626525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,65536,0.0849555532137553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,65536,0.0456026660071479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,65536,0.07819733354780409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,16384,0.024726221958796184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,16384,0.02390844457679325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,12288,0.011699555648697747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,12288,0.01973155637582143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,10240,0.010889777706729041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,10240,0.017768000562985737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,8192,0.01516533394654592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,8192,0.010777778095669217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,8192,0.015406222807036506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,7168,0.010621333287821876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,7168,0.01440533333354526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,6144,0.009945777555306753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,12288,0.019663110375404358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,10240,0.017401778035693698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,6144,0.013444444371594323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,5120,0.01035288886891471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,5120,0.009391999906963771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,5120,0.011881777809725868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,4096,0.008255999949243333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,4096,0.009547555612193214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,7168,0.013797333670987023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,3584,0.00848355558183458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,6144,0.011844444606039258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,4096,0.008944888909657797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,3584,0.008604444563388824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,3584,0.00920800036854214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,3072,0.007568000091446771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,3072,0.008632000121805403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,2560,0.007259555160999298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,2560,0.006521777974234686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,2560,0.00812355594502555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,2048,0.006603555546866522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,2048,0.0057751110030545135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,2048,0.0074746666683091064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,3072,0.007830222447713217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,1536,0.006120000034570694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,1536,0.005051555732885997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,1536,0.007054222126801808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,1024,0.005473777651786804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,1024,0.004318222403526306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,1024,0.006480000085300869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,768,0.0052133335007561584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,768,0.00398577791121271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,768,0.00629066675901413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,512,0.0036346668170558084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,512,0.006024888820118374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,256,0.004550222307443619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,256,0.0032231110251612137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,256,0.005793777604897817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,128,0.002970666728085942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,128,0.005601777798599667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,64,0.00281866660548581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,64,0.005649777750174205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,512,32,0.0028488888508743713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,512,32,0.005724444571468566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,512,0.004831999954250124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,512,128,0.004371555729044808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,65536,0.03139999839994643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,16384,0.024303111765119765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,65536,0.0786844425731235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,16384,0.011532444092962476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,16384,0.023805333508385554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,12288,0.010252444280518426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,65536,0.08457155360115899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,12288,0.019508444600635104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,10240,0.009752000371615091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,10240,0.017533333765135873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,8192,0.010075555907355415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,8192,0.015062222878138224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,7168,0.009315555294354757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,7168,0.01420177850458357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,6144,0.011393778026103973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,12288,0.01953066719902886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,6144,0.008967111508051554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,10240,0.017128000656763714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,8192,0.014446222119861178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,6144,0.0125591109196345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,5120,0.00999111102686988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,5120,0.008079999850855933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,7168,0.01292888902955585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,5120,0.01074577785200543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,4096,0.007187555233637492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,4096,0.009483555952707926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,3584,0.00832444429397583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,3584,0.007369777394665613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,3584,0.009080000221729279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,3072,0.007656888829337225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,3072,0.007055111229419708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,3072,0.008519111408127679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,2560,0.007146666447321574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,2560,0.00628977765639623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,2560,0.007920888562997183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,2048,0.0064186664919058485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,2048,0.005605333381228977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,2048,0.007394666473070781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,1536,0.004906666775544484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,1536,0.0069155556460221606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,1024,0.005315555466545953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,4096,0.008830221990744272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,1024,0.00423911131090588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,1024,0.0064311110311084324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,1536,0.006055111272467508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,768,0.005019555489222209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,768,0.00397777764333619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,768,0.006236444330877728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,512,0.0047360000511010485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,512,0.003645333151022593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,512,0.005934222290913264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,256,0.004420444369316101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,256,0.0032559999575217566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,256,0.005691555639108022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,128,0.0029564443975687027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,128,0.005535111245181825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,64,0.0028000000036425064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,64,0.005531555662552516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,256,32,0.0028613333900769553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,256,32,0.005583111196756363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,65536,0.0842755569352044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,65536,0.021320889393488567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,65536,0.07936710781521268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,16384,0.010435555544164447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,16384,0.023864888482623633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,12288,0.009362666971153682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,12288,0.01958488921324412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,10240,0.016395555602179635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,256,128,0.004272888931963178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,10240,0.008819555242856344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,16384,0.024305777417288885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,10240,0.01754577789041731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,8192,0.013968000809351603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,8192,0.007617777420414819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,8192,0.014246222045686511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,7168,0.0074480002125104266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,7168,0.012958221965365939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,12288,0.019368888603316415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,6144,0.006763555523422029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,6144,0.011671110987663269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,5120,0.0071937781241205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,5120,0.010642666783597736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,4096,0.008754666480753157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,4096,0.006721777634488211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,4096,0.009588444398509132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,3584,0.008273777862389883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,3584,0.007187555233637492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,3584,0.009104889300134447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,3072,0.00686488880051507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,7168,0.012640000217490725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,6144,0.01129066695769628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,3072,0.008461332983440822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,2560,0.006151111175616582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,2560,0.00796088907453749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,2048,0.0063573333124319715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,2048,0.005532444351249271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,2048,0.007366221812036302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,1536,0.005936000082227919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,1536,0.004810666872395409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,1536,0.007034666836261749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,1024,0.0052328887912962176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,1024,0.004101333518822988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,1024,0.0064479998416370815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,768,0.004960888789759742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,768,0.0037253333462609183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,768,0.006253333141406377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,512,0.004625777817434735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,512,0.003391111062632667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,512,0.005944889038801193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,256,0.004415110995372136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,256,0.003063999944263034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,256,0.005669333454635408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,128,0.0042071110672420925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,128,0.0028693332440323303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,128,0.005538666827811136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,64,0.002712888850106133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,64,0.005526222288608551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,128,32,0.0027599999060233435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,128,32,0.005597333527273602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,65536,0.01796444422668881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,65536,0.07986666758855183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,16384,0.00867377801073922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,16384,0.023752000596788194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,5120,0.009927999642160203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,3072,0.007598222129874759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,12288,0.007871111234029135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,12288,0.01944177846113841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,10240,0.007714666426181793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,10240,0.017328000730938382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,256,128,2560,0.0070497774415545994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,8192,0.007224000162548489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,8192,0.013974222871992322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,7168,0.006862222320503659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,7168,0.012434666355450949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,6144,0.006605333338181178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,6144,0.011503111157152386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,5120,0.007015110717879401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,5120,0.010584000084135266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,4096,0.006592000110281839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,4096,0.009360888765917884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,3584,0.007073777417341868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,3584,0.00889866633547677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,3072,0.006684444430801604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,3072,0.008445333275530074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,2560,0.006045333213276333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,2560,0.007909333540333642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,2048,0.005416000054942236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,2048,0.007297777467303806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,1536,0.0047066667013698155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,1536,0.006884444504976273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,1024,0.004037333445416557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,1024,0.0063777777055899305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,768,0.003648888733651903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,768,0.006161777923504512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,512,0.003330666571855545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,512,0.0058782220714622075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,256,0.003007111036115222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,256,0.005647111270162795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,128,0.0028328889360030494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,128,0.005489777773618698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,64,0.0026444445053736367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,64,0.005511111269394557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,64,32,0.0026817777090602448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,64,32,0.005493333356248007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,65536,0.016935111747847665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,65536,0.07950577470991346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,16384,0.008851555486520132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,16384,0.023631110787391663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,12288,0.00739911115831799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,12288,0.019254222512245178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,10240,0.007261333366235097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,10240,0.017087999317381117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,8192,0.0068897778789202375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,8192,0.013745778136783175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,7168,0.006670222100284364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,7168,0.012307555311255984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,6144,0.006493333313200209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,6144,0.011238222320874533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,5120,0.00686488880051507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,5120,0.010449777874681683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,4096,0.0064488889442549805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,4096,0.0092257774538464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,3584,0.0069066666894488865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,3584,0.008861333131790161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,3072,0.006608888920810487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,3072,0.008449777960777283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,2560,0.006108444597986009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,2560,0.007838221887747446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,2048,0.005400889035728242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,2048,0.007240888973077138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,1536,0.0047280001971456735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,1536,0.006843555718660355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,1024,0.00406666679514779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,1024,0.006351111249791251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,768,0.0037066667444176147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,768,0.006123555617200003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,512,0.003350222276316749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,512,0.0058293334311909145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,256,0.0030248889492617715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,256,0.005607999861240387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,128,0.0028000000036425064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,128,0.00545866663257281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,64,0.0026586666289303037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,64,0.005435555759403441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,256,32,32,0.0026311110705137253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,256,32,32,0.005477333234416113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,16384,0.6491840150621202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,12288,0.49659024344550234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,12288,0.35955378744337296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,16384,0.4709208806355794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,12288,0.4284817907545302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,16384,0.606952879163954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,10240,0.40059733390808105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,8192,0.2371679941813151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,10240,0.35222933027479386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,10240,0.2926657729678684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,8192,0.3323440021938748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,7168,0.29213423199123806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,7168,0.20350578096177843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,8192,0.2717493375142415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,6144,0.24800178739759657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,7168,0.24381865395439994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,5120,0.1497839954164293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,5120,0.21269422107272676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,6144,0.19528888331519234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,4096,0.16824977927737764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,4096,0.12364178233676487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,5120,0.1462497843636407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,4096,0.12138488557603623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,6144,0.18106933434804282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,3584,0.14979378382364908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,3584,0.10786755879720052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,3072,0.13220266501108804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,3072,0.09431466791364883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,2560,0.1114399962955051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,2560,0.07798133293787639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,2560,0.08140355348587036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,2048,0.06599555412928264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,2048,0.09256088733673096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,2048,0.06790755854712592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,1536,0.04974666568968031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,3584,0.11226133505503337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,1536,0.07080088721381293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,3072,0.09363022115495469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,1536,0.053782221343782216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,1024,0.03577333357599046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,1024,0.05158844590187073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,1024,0.04403555393218994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,768,0.02980533242225647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,768,0.041653331783082746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,768,0.038198222716649376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,512,0.025452444950739544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,512,0.032447109619776406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,512,0.03418666786617703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,256,0.01945244438118405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,256,0.023583999938435022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,256,0.03050666716363695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,128,0.018225777480337355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,128,0.028804444604449805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,64,0.015782222151756287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,64,0.02994577752219306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,65536,32,0.01629244453377194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,65536,128,0.015423110789722867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,65536,32,0.03127822279930115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,65536,0.6418595314025879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,65536,0.6547022395663792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,65536,0.4739057752821181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,16384,0.12457066112094456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,16384,0.16212356090545654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,12288,0.12121600574917263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,12288,0.09447555409537421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,16384,0.14672088623046875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,12288,0.0921857754389445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,10240,0.10399644242392646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,10240,0.08003377914428711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,10240,0.08202577961815728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,8192,0.08197244670655993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,8192,0.06294933292600843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,8192,0.06198400259017944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,7168,0.07520177629258898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,7168,0.055234667327668935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,6144,0.04849777619043986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,6144,0.0631493330001831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,7168,0.05801244576772054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,6144,0.0464053319560157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,5120,0.05388266510433621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,5120,0.04052177733845181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,4096,0.043678220775392324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,4096,0.033463110526402794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,3584,0.03143289022975498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,5120,0.04237599836455452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,3584,0.03829333186149597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,4096,0.0337066650390625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,3584,0.030760000149408977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,3072,0.027341332700517442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,3072,0.033886220720079206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,3072,0.02737066646416982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,2560,0.02914311157332526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,2560,0.024176888995700415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,2560,0.024780443973011438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,2048,0.02418488926357693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,2048,0.02090666691462199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,2048,0.02200711104604933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,1536,0.017804443836212158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,1536,0.019501333435376484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,1536,0.018400000201331247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,1024,0.014440889159838358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,1024,0.013825777504179211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,1024,0.0155102229780621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,768,0.012882666455374824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,768,0.011316444310877057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,768,0.014135999812020196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,512,0.00907555553648207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,512,0.013204444613721637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,256,0.007119999991522894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,256,0.010479999913109673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,128,0.008555555509196388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,128,0.005121777868933148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,128,0.009754666851626502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,64,0.004479111068778568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,64,0.010126222338941362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,16384,32,0.004896000027656555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,16384,32,0.011752888560295105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,65536,0.3155493206448025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,16384,0.08258222209082709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,16384,0.12609155972798666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,512,0.011271110839313932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,65536,0.48256535000271267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,16384,0.11881777975294326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,12288,0.061695112122429736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,12288,0.09350755479600693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,65536,0.5411778026156956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,16384,256,0.009945777555306753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,12288,0.07274578015009563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,10240,0.05302310983339945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,10240,0.06515110863579644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,10240,0.08277155293358697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,8192,0.06279466549555461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,8192,0.042985777060190834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,8192,0.05277599891026815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,7168,0.05630489190419515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,7168,0.044659554958343506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,6144,0.03372622198528714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,6144,0.04830133252673679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,6144,0.039064000050226845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,5120,0.02940444482697381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,5120,0.04122133387459649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,5120,0.03388799892531501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,4096,0.02500533395343357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,4096,0.03384088807635837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,7168,0.03831200136078729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,4096,0.029176000091764662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,3584,0.023022222850057814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,3584,0.030208001534144085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,3584,0.026131555438041687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,3072,0.02102577851878272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,3072,0.026526222626368206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,3072,0.02365600069363912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,2560,0.018810666269726224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,2560,0.02238044473859999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,2560,0.021333333518770006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,2048,0.01903288894229465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,2048,0.018964444597562153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,1536,0.014733332726690503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,1536,0.015966221690177917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,1024,0.012621333201726278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,1024,0.010943111446168689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,1024,0.013835555149449242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,768,0.009310222334331935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,2048,0.01685066686736213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,1536,0.014671110444598727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,768,0.012842666771676807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,768,0.011121778024567498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,512,0.010138666464222802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,512,0.00774133370982276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,512,0.01180266671710544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,256,0.008648888932334052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,256,0.005663110978073544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,256,0.009538667069541084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,12288,128,0.00665244460105896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,128,0.00452622233165635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,128,0.008987555901209513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,64,0.004023999803596073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,64,0.00925511121749878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,12288,32,0.004426666845877965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,12288,32,0.010265777508417765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,65536,0.29828177558051217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,65536,0.4164880116780599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,65536,0.43929065598381895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,16384,0.07376355595058866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,16384,0.1214560005399916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,16384,0.09136799971262614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,12288,0.09046577745013767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,12288,0.05883466535144382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,12288,0.05585600270165337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,10240,0.04760266674889458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,10240,0.07720977730221219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,10240,0.05217422379387749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,8192,0.03923199905289544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,8192,0.061632891496022545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,8192,0.04215288824505276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,7168,0.0349333319399092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,7168,0.05406399899058872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,7168,0.037123554282718234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,6144,0.03077955709563361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,6144,0.044237332211600415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,6144,0.033392889632119074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,5120,0.03805155555407206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,5120,0.028871999846564397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,4096,0.022840888963805303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,4096,0.0309804446167416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,4096,0.02496888902452257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,5120,0.026711111267407734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,3584,0.021171554923057556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,3584,0.029282665914959375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,3584,0.022525333695941504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,3072,0.019103111492262948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,3072,0.024354666471481323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,3072,0.020563556088341605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,2560,0.017322666115230985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,2560,0.021037333541446265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,2560,0.018498667412334018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,2048,0.015352000792821249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,2048,0.018730666902330186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,2048,0.016513778103722464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,1536,0.013345777988433838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,1536,0.015211555692884656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,1536,0.014142221874660917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,1024,0.010654222634103564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,1024,0.01236355553070704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,768,0.010120888551076254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,768,0.009113777842786577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,768,0.011667555405033959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,512,0.009306666751702627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,512,0.007658667034573025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,1024,0.01128799964984258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,512,0.010004444254769219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,256,0.007335999773608313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,256,0.0053342220683892565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,256,0.00868266655339135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,128,0.0045422220395671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,10240,128,0.006085333310895496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,128,0.0083137775460879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,64,0.003998222036494149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,10240,32,0.004505777938498391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,64,0.008555555509196388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,10240,32,0.00941777808798684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,65536,0.32196177376641166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,65536,0.3327137894100613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,16384,0.05991022454367744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,16384,0.0926097763909234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,16384,0.0722586645020379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,65536,0.23185422685411242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,12288,0.04557333389918009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,12288,0.07317155599594116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,12288,0.04761955473158094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,10240,0.040215111441082425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,10240,0.05609333515167236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,10240,0.0403422216574351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,8192,0.032560888263914324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,8192,0.046451555358039014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,8192,0.03413955701722039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,7168,0.029102222787009344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,7168,0.039572444227006696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,7168,0.030664887693193223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,6144,0.035001777940326266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,6144,0.027535999814669292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,5120,0.022823999325434368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,5120,0.029696888393825952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,5120,0.0239982224173016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,4096,0.024664888779322307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,4096,0.02051999999417199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,6144,0.025923555095990498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,4096,0.019706666469573975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,3584,0.02199911077817281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,3584,0.018720888429217868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,3072,0.019528888993793063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,3072,0.01700444519519806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,2560,0.015153777268197803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,2560,0.015592889653311836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,2560,0.01569422251648373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,3584,0.018395556343926322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,2048,0.01352355546421475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,2048,0.013000888956917657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,2048,0.014093332820468478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,1536,0.01202311118443807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,3072,0.01667555504375034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,1536,0.010556444525718689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,1536,0.012300444145997366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,1024,0.008411555654472774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,1024,0.011148444480366178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,768,0.009279111193286048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,768,0.007446222007274628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,768,0.010042666561073726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,512,0.0064311110311084324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,512,0.008376888930797577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,1024,0.010266666611035665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,256,0.00582755563987626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,256,0.00453599997692638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,256,0.007770666645632849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,512,0.007858666280905405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,8192,128,0.005361777626805835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,128,0.003965333518054751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,128,0.0074471111098925276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,64,0.003640000190999773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,64,0.007655999726719326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,8192,32,0.00398488880859481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,8192,32,0.00813333359029558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,65536,0.29943643675910103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,65536,0.2598675621880425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,65536,0.3201955424414741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,16384,0.08228977521260579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,16384,0.06521422333187528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,12288,0.04725066820780436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,12288,0.06535555256737603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,12288,0.04600977897644043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,16384,0.06435822116004096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,10240,0.040639999839994646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,10240,0.050154666105906166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,10240,0.039656000004874334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,7168,0.0295448899269104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,8192,0.033144887950685285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,6144,0.030244443151685927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,8192,0.041133334239323936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,8192,0.03365600109100342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,7168,0.03470844361517165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,6144,0.02699022160636054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,5120,0.025940444734361436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,5120,0.02252799934811062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,5120,0.023815999428431194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,4096,0.019090667366981506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,4096,0.021878222624460857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,7168,0.028696888022952612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,6144,0.025784000754356384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,4096,0.020035554965337116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,3584,0.017849778135617573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,3584,0.019719999697473314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,3584,0.018380444910791185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,3072,0.017323555217848886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,3072,0.01683911184469859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,2560,0.014825777875052558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,2560,0.014984889162911309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,2560,0.015586665934986539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,2048,0.013103110922707452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,3072,0.016065778003798593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,2048,0.01261333293384976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,2048,0.014053333136770459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,1536,0.01184088902340995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,1536,0.010617777705192566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,1536,0.012166221936543783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,1024,0.010407110883129967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,1024,0.008411555654472774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,1024,0.010967111421955956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,768,0.00906577789121204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,768,0.007254222200976477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,768,0.00927022182279163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,512,0.007558222446176741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,512,0.006312888943486744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,512,0.008377778033415476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,256,0.006156444549560547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,256,0.0046675557063685525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,256,0.00775111135509279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,128,0.0041555555330382455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,128,0.007391110890441471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,64,0.00360000009338061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,64,0.007687999970383114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,7168,128,0.005685333162546158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,7168,32,0.003809777812825309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,7168,32,0.008106666306654612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,65536,0.2467279964023166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,16384,0.05046222276157803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,16384,0.07180355654822455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,65536,0.2898346583048503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,16384,0.062353776560889355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,65536,0.1918755504820082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,12288,0.0565626687473721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,12288,0.044975999328825206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,10240,0.048594666851891406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,10240,0.039600888888041176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,8192,0.04181955589188469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,8192,0.03299199872546726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,10240,0.03429777754677667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,12288,0.03940266701910231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,7168,0.025895110434956018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,7168,0.030368000268936157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,7168,0.029532445801628962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,6144,0.026908444033728704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,8192,0.028334223561816748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,6144,0.02674755619631873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,6144,0.022912888063324824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,5120,0.020267556111017864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,5120,0.022823999325434368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,5120,0.023543111152119104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,4096,0.017477333545684814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,4096,0.021021333005693223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,4096,0.019685332973798115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,3584,0.016164445214801364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,3584,0.018617777360810172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,3584,0.018170666363504197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,3072,0.015543111496501498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,3072,0.016688888271649677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,3072,0.014979556202888489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,2560,0.01370488852262497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,2560,0.015210666590266757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,2048,0.012686221963829465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,2560,0.013834666874673633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,2048,0.011532444092962476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,2048,0.013910222384664746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,1536,0.01107022249036365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,1536,0.009683555199040307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,1536,0.012117333710193634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,1024,0.007600888609886169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,1024,0.010399111443095738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,768,0.008255999949243333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,768,0.006835555450783835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,768,0.008927110996511247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,512,0.005640000104904175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,1024,0.010191111101044549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,512,0.006685333533419504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,512,0.008176888856622908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,256,0.005616000129116907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,256,0.004302222281694412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,256,0.007782222496138678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,128,0.0038542221817705366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,128,0.007305777735180325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,64,0.0034808889031410217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,64,0.007538666327794393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,6144,32,0.003571555432346132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,6144,32,0.007747555772463481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,65536,0.20959111054738364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,16384,0.0614853302637736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,16384,0.06006311045752632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,65536,0.25124621391296387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,65536,0.21610044108496773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,6144,128,0.005249777601824866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,16384,0.04821333289146423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,12288,0.04920533299446106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,12288,0.033702221181657575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,12288,0.04599377844068739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,10240,0.03720088799794515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,10240,0.03968266646067301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,10240,0.030192888445324365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,8192,0.033913777934180364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,8192,0.024871110916137695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,7168,0.027837332752015855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,8192,0.03133599956830343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,7168,0.02283199959331089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,6144,0.0240657776594162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,6144,0.024482667446136475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,7168,0.027808000644048054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,6144,0.02071199980047014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,5120,0.020968000094095867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,5120,0.021274665991465252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,5120,0.018603555030292936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,4096,0.017664889494578045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,4096,0.017850667238235474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,4096,0.016181333197487723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,3584,0.016224000189039443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,3584,0.01715288890732659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,3584,0.014850666125615438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,3072,0.014431110686726041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,3072,0.01386666629049513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,3072,0.013808888693650564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,2560,0.013359111216333179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,2560,0.012882666455374824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,2560,0.012905778156386482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,2048,0.011392888923486074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,2048,0.011477332976129321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,2048,0.011785777906576792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,1536,0.0101742222905159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,1536,0.008536888493431939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,1536,0.010416888528399997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,1024,0.0070088886552386815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,1024,0.008554666406578487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,1024,0.00850133349498113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,768,0.007287110719415877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,768,0.006246222390068903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,768,0.007742221984598372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,512,0.006471111128727595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,512,0.0047439999050564235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,512,0.007221333682537079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,256,0.004023111114899318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,256,0.006804444309737947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,128,0.003699555579158995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,128,0.006595555692911148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,256,0.005677333308590784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,64,0.0033440000067154565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,64,0.006892444358931647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,5120,32,0.0034640000926123727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,5120,128,0.0052737775776121355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,5120,32,0.007125332951545715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,65536,0.1682577795452542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,65536,0.15118222766452366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,65536,0.19112711482577852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,16384,0.04318755533960131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,16384,0.05057511064741346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,16384,0.043649779425727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,12288,0.04197777642144097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,12288,0.0322675539387597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,12288,0.03365600109100342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,10240,0.029000888268152874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,10240,0.04220444295141432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,10240,0.02903555499182807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,8192,0.02387822171052297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,8192,0.03754488958252801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,8192,0.024674667252434626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,7168,0.02168888847033183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,7168,0.029394666353861492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,7168,0.0225608895222346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,6144,0.026555554734336004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,6144,0.020121778051058452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,5120,0.017443555924627516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,5120,0.02049333353837331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,6144,0.019451555278566148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,5120,0.0181031111213896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,4096,0.017981333865059745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,4096,0.01588800052801768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,3584,0.014053333136770459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,3584,0.015788444214397006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,4096,0.015172445111804538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,3584,0.014602666099866232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,3072,0.012905778156386482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,3072,0.01294133315483729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,3072,0.013550221920013428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,2560,0.011896889242861005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,2560,0.011240888800885944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,2560,0.01273333364062839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,2048,0.010635555618339114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,2048,0.009455111291673448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,2048,0.011576889289749993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,1536,0.008984888593355814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,1536,0.008206222620275285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,1536,0.009974222216341231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,1024,0.006970666348934174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,1024,0.0064382221963670515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,1024,0.008247999681366814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,768,0.006037333359320958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,768,0.005324444423119227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,768,0.007762666377756331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,512,0.005568000177542369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,512,0.004559111264016894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,512,0.007264888948864407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,256,0.00388355553150177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,256,0.0068408888247278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,128,0.003568888952334722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,128,0.006569777925809224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,64,0.0033537778589460584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,64,0.0068986668354935115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,4096,32,0.003391111062632667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,4096,32,0.006907555378145642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,256,0.005148444324731827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,65536,0.15833155314127603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,16384,0.04212177793184916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,16384,0.0469395551416609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,65536,0.1833040051990085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,16384,0.04231466518508064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,12288,0.03276622295379639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,12288,0.037652442852656044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,12288,0.0322053333123525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,4096,128,0.004818666726350784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,65536,0.14869688616858587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,10240,0.02824266751607259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,10240,0.03289777702755398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,10240,0.028400000598695543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,8192,0.02440533373090956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,8192,0.02401777770784166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,7168,0.021614222062958613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,7168,0.021951110826598272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,8192,0.023216888308525085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,6144,0.018963555494944256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,7168,0.021159110797776118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,6144,0.020068445139461093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,5120,0.016921778519948322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,5120,0.016587555408477783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,5120,0.017992888887723286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,3584,0.012666666673289405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,6144,0.01917777789963616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,3072,0.01145155562294854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,4096,0.018413333429230582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,4096,0.015484443969196744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,3584,0.01460800071557363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,3072,0.012715555727481842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,3072,0.013575110998418597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,2560,0.011519999967681037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,2560,0.009552000297440423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,2560,0.012562666502263812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,2048,0.008433777425024245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,2048,0.011643555429246692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,1536,0.008608000146018134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,4096,0.014821334017647637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,3584,0.013740444348918067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,2048,0.010409778190983666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,1536,0.007221333682537079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,1536,0.009429333110650381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,1024,0.006086222413513396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,1024,0.008185777399275038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,768,0.006032889087994893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,768,0.004957333207130432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,768,0.007696000238259633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,1024,0.006585777633719974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,512,0.005484444399674733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,512,0.0041759999261962045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,512,0.007298666569921706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,256,0.005065777649482091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,256,0.0036640001667870414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,256,0.006795555353164673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,128,0.0033599999215867785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,128,0.006530666516886816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,64,0.00314666661951277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3584,128,0.004779555731349521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,64,0.006739555547634761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3584,32,0.0032293332947625052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3584,32,0.006846222198671765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,65536,0.209626661406623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,65536,0.1313306623035007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,65536,0.16107822789086237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,16384,0.05369244350327385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,16384,0.04172799984614054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,16384,0.039600001441107854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,12288,0.04080533319049411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,12288,0.03191377719243368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,12288,0.03122933374510871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,10240,0.03445511062939962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,10240,0.02753244505988227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,10240,0.02760266595416599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,8192,0.028991109795040552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,8192,0.02276533345381419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,8192,0.023269333773189124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,7168,0.025145777397685583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,7168,0.02030933399995168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,7168,0.021287999219364587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,6144,0.02186577849917942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,6144,0.017774222625626456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,6144,0.019381332728597853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,5120,0.018032888571421306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,5120,0.016576889488432143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,5120,0.017704889178276062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,4096,0.016011555989583332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,4096,0.012792000340090858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,3072,0.013169777890046438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,4096,0.015255999233987598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,3584,0.014768888552983602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,3584,0.01148977792925305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,3584,0.01423822177780999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,3072,0.013192888763215808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,3072,0.012074666718641916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,2560,0.011801777614487542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,2560,0.01057866629627016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,2560,0.012146666646003723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,2048,0.010101333260536194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,2048,0.008910222185982598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,2048,0.010869333313571082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,1536,0.007849777738253275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,1536,0.008838222258620793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,1024,0.006707555717892117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,1024,0.005952888892756567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,1024,0.007962666451931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,1536,0.008353778057628209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,768,0.005143110950787862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,768,0.006232888748248418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,768,0.0075937774446275495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,512,0.004401777767472797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,512,0.0070977773931291364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,256,0.005283555636803309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,256,0.0036355555057525635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,256,0.006789333290523953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,128,0.00500177757607566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,128,0.0033555556502607134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,128,0.006747555401590135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,3072,512,0.005758222192525864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,64,0.0031351111829280853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,64,0.006683555742104848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,3072,32,0.003216000066863166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,3072,32,0.006774222271309958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,65536,0.09925778044594659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,65536,0.11169600486755371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,65536,0.10834755500157674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,16384,0.028543111350801256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,16384,0.03749688797526889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,16384,0.026561778452661302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,12288,0.030788444810443457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,12288,0.02164799968401591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,12288,0.022755554980701868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,10240,0.026857778429985046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,10240,0.020311110549502902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,10240,0.019451555278566148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,8192,0.02369600037733714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,8192,0.01686755485004849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,8192,0.01683555543422699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,7168,0.01551999979548984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,7168,0.016755556066830952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,7168,0.015607110328144498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,6144,0.014217777384652032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,6144,0.016377778516875375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,6144,0.01443733274936676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,5120,0.012950222525331708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,5120,0.01278222186697854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,5120,0.013362666798962487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,4096,0.0114702218108707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,4096,0.01274400038851632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,4096,0.011782222323947482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,3584,0.01092355532778634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,3584,0.011037333144081963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,3584,0.01131377783086565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,3072,0.010115555591053432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,3072,0.009151111046473185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,3072,0.010684444672531553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,2560,0.008657777474986183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,2560,0.01000533335738712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,2048,0.008091555701361762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,2048,0.007649777664078607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,2048,0.008792888787057664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,2560,0.009654222263230218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,1536,0.006962666908899943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,1536,0.006664000037643645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,1536,0.007429333196745978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,1024,0.0052764444715446895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,1024,0.006913777854707506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,768,0.005611555443869696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,768,0.004508444418509801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,768,0.006703111032644908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,1024,0.005945777727497949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,512,0.003983111017280155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,512,0.0063946665161185795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,256,0.003564444267087512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,256,0.006048000107208888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,512,0.0052693333062860705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,256,0.004864888886610667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2560,128,0.0046364445653226645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,128,0.00328177772462368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,128,0.005807111246718301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,64,0.003040888864133093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,64,0.006059555543793573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2560,32,0.0031537777847713898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2560,32,0.00619555554456181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,65536,0.09133244223064846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,65536,0.09012799792819554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,65536,0.09592355622185601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,16384,0.03053599927160475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,16384,0.025940444734361436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,12288,0.02936977810329861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,16384,0.0278479986720615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,12288,0.022500443789694045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,12288,0.02110044492615594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,10240,0.01963911122745938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,10240,0.025432889660199482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,10240,0.018960888187090557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,8192,0.01642311116059621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,8192,0.023095111052195232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,8192,0.016492444607946608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,7168,0.01531733406914605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,7168,0.01550222271018558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,7168,0.015357333752844067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,6144,0.01389422184891171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,6144,0.013744889034165276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,6144,0.014256000518798828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,5120,0.012221333053376941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,5120,0.013074666261672974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,4096,0.011195555329322815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,5120,0.012493333054913415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,4096,0.010695999695195092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,4096,0.011634666886594562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,3584,0.010620444185203977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,3584,0.009698666632175446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,3584,0.011085333095656501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,3072,0.009361777868535783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,3072,0.00887822194231881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,3072,0.01034755590889189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,2560,0.007975111405054728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,2560,0.00903911143541336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,2048,0.0074622225430276655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,2048,0.007162666983074612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,2048,0.007901333272457123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,1536,0.006636444479227066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,1536,0.006309333360857434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,2560,0.008565333154466417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,1536,0.007309333317809635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,1024,0.005807111246718301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,1024,0.004831999954250124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,1024,0.006793777975771163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,768,0.004410666724046071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,768,0.006582222051090664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,512,0.005155555489990446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,512,0.003968888686762916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,512,0.006365333166387346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,768,0.005437333136796951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,256,0.004791111167934206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,256,0.0034826666944556763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,256,0.005998222364319696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,2048,128,0.004581333448489507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,128,0.0032586666444937387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,128,0.005817777580685086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,64,0.003026666740576426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,64,0.005920889063013925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,2048,32,0.0030862221287356485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,2048,32,0.006072000082996156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,65536,0.09416977564493816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,65536,0.07425422138637967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,65536,0.08277510934405856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,16384,0.027445332871543035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,16384,0.024865777956114873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,12288,0.022545778089099463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,16384,0.027521777484152053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,12288,0.020242666204770405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,12288,0.021935999393463135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,10240,0.019309333629078336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,10240,0.019851555426915485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,10240,0.018188445104493033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,8192,0.01682933337158627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,8192,0.01575644479857551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,7168,0.015383111106024848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,7168,0.01483022173245748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,6144,0.014231999715169271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,6144,0.013752000199423896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,8192,0.016189333465364244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,5120,0.0123511114054256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,5120,0.013106666505336761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,5120,0.012785777449607849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,7168,0.015034667319721647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,4096,0.010919999745157031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,6144,0.013615999784734515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,4096,0.011933333343929715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,4096,0.011206222077210745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,3584,0.010058666268984476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,3584,0.011310222248236338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,3584,0.010424888796276517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,3072,0.00923733330435223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,3072,0.008681777450773451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,3072,0.00943199959066179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,2560,0.007920000288221572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,2560,0.00849511143234041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,2048,0.007091555330488417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,2560,0.00814488861295912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,2048,0.0069164443347189165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,2048,0.007688889073001013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,1536,0.006548444430033366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,1536,0.005585777676767773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,1536,0.00721244431204266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,1024,0.005723555468850666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,1024,0.004638222356637319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,1024,0.006723555425802867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,768,0.005377777748637729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,256,0.003419555516706573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,768,0.004320888883537716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,768,0.006537777682145436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,512,0.0038293335172865125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,512,0.00628266649113761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,256,0.0047048889100551605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,256,0.005941333456171884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,128,0.004477333277463913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,128,0.003132444495956103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,128,0.005757333503829108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,64,0.0029760001020299066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,64,0.005853333406978183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1536,512,0.005056000004212062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1536,32,0.003043555551105075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1536,32,0.00592533333433999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,65536,0.0521360006597307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,65536,0.16367199685838488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,65536,0.07731378078460693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,16384,0.04402933186954922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,16384,0.019493333167499967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,16384,0.023916444844669763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,12288,0.03370133373472426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,12288,0.01590577761332194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,12288,0.01975200076897939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,10240,0.027775999572541978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,10240,0.01513155632548862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,10240,0.017700443665186565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,8192,0.02379200028048621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,8192,0.013733333183659447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,8192,0.01536266671286689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,5120,0.015594666202863058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,7168,0.02106844385464986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,7168,0.012496000362767113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,7168,0.014427555931939019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,6144,0.01847644481394026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,6144,0.011608888705571493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,6144,0.013564444250530668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,5120,0.010846222440401712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,5120,0.012452444268597497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,4096,0.013045333325862885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,4096,0.009939555492666032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,4096,0.010241777532630498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,3584,0.00905511114332411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,3584,0.009901333186361525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,3072,0.010300444232092964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,3072,0.00832266691658232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,3072,0.00886933339966668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,3584,0.01182666669289271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,2560,0.009356444080670675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,2560,0.007463110817803278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,2560,0.00812977800766627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,2048,0.0064426664676931165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,2048,0.007546666595670912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,1536,0.007263110743628607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,1536,0.005231110999981563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,1536,0.007111111448870764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,1024,0.004426666845877965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,1024,0.006641777853171031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,2048,0.008043555749787224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,768,0.0058204444746176405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,768,0.004192000048028098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,1024,0.006167999986145232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,768,0.00645955569214291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,512,0.0053546664615472155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,512,0.0036951113078329298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,512,0.006131555471155379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,256,0.003335111050142182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,64,0.00572533326016532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,256,0.0058515556156635284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,128,0.004688888788223267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,128,0.003006222140457895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,1024,256,0.004912000149488449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,128,0.005705777969625261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,64,0.0028568889117903183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,1024,32,0.0028880000528362063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,1024,32,0.005741333381997214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,65536,0.04278666774431864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,65536,0.13666133085886636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,65536,0.07513511180877686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,16384,0.03879200087653266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,16384,0.01649599936273363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,16384,0.02370311154259576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,12288,0.030839999516805012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,12288,0.014009777042600842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,12288,0.01959022217326694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,10240,0.026199110680156287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,10240,0.01347644461525811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,10240,0.01759822169939677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,8192,0.011935999823941125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,8192,0.015277332729763456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,7168,0.02030666669209798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,7168,0.011400889191362592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,7168,0.014440889159838358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,6144,0.016189333465364244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,6144,0.010597333312034607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,6144,0.013280889226330651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,5120,0.01460088955031501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,5120,0.009992000129487779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,3584,0.008786666724416945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,5120,0.011993777420785693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,8192,0.022279999322361414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,4096,0.009310222334331935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,4096,0.009782222410043081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,3584,0.011387555963463254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,3584,0.009212444225947062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,3072,0.00998133338159985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,3072,0.007834666305118138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,4096,0.012368888490729861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,3072,0.00869955536392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,2560,0.006855999843941794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,2560,0.008119111259778341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,2048,0.007663999994595845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,2048,0.005746666755941179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,2048,0.007500444849332173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,1536,0.005125333451562458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,1536,0.007054222126801808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,1024,0.005977777971161737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,1024,0.004330666528807746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,1024,0.006585777633719974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,2560,0.00867733359336853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,768,0.005522666705979242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,1536,0.006808888994985157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,768,0.003999999827808804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,768,0.0063822223908371395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,512,0.005299555758635203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,512,0.0036159998012913596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,512,0.006062222023804982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,256,0.0032195556494924757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,256,0.00583022211988767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,128,0.004573333180612988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,128,0.0029680000411139596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,128,0.005689777847793367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,64,0.002826666666401757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,64,0.005758222192525864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,768,32,0.002883555574549569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,768,32,0.005745777653323279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,768,256,0.004831999954250124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,65536,0.032850666178597346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,65536,0.08035733302434285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,65536,0.07582933372921415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,16384,0.014038221703635322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,16384,0.02348800003528595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,12288,0.01900533338387807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,12288,0.011624000138706632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,12288,0.019317333896954853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,10240,0.010760000182522668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,10240,0.017430222696728177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,8192,0.01401422255569034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,8192,0.009921777579519484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,16384,0.02312711046801673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,8192,0.015243555108706156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,10240,0.01696266730626424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,7168,0.00925511121749878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,7168,0.014192889134089151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,6144,0.011168888873524137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,7168,0.013020444247457715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,6144,0.008827555510732863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,6144,0.012477333347002665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,5120,0.009959110783206092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,5120,0.011241777903503843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,4096,0.009134222235944536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,4096,0.009518221848540837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,5120,0.009855111440022787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,3584,0.008221333225568136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,4096,0.008629333641793992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,3584,0.009135111338562435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,3072,0.007333333293596904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,3072,0.008603555460770925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,2560,0.007033777733643849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,3584,0.008226667013433244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,2560,0.0063795554969045855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,2560,0.008032889001899296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,2048,0.006511111226346757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,3072,0.007707555260923173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,2048,0.005666666560702854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,2048,0.007429333196745978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,1536,0.006047111004590988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,1536,0.004976888911591636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,1536,0.0069644442862934535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,1024,0.005379555539952383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,1024,0.004265777766704559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,1024,0.006551110910044775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,768,0.003887999802827835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,768,0.006333333336644703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,512,0.004755555755562252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,768,0.005025777965784073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,512,0.003540444291300244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,512,0.005991111199061076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,256,0.004471999903519948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,256,0.003134222287270758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,256,0.0058044443527857465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,512,128,0.004334222111437055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,128,0.0028702221396896574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,128,0.005640888793600931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,64,0.0027662221756246355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,64,0.005664888769388199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,512,32,0.0028408887899584244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,512,32,0.005685333162546158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,65536,0.0796568857298957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,65536,0.025814221964942083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,65536,0.07623644669850667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,16384,0.022895110978020564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,16384,0.010187555518415239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,16384,0.02344888945420583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,12288,0.018580444984965853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,12288,0.010488000180986194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,12288,0.019310222731696237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,10240,0.01610488858487871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,10240,0.009740444521109263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,10240,0.017320000463061862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,8192,0.013270222478442721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,8192,0.008877333667543199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,8192,0.01423111061255137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,7168,0.012064889073371887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,7168,0.008007999923494127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,7168,0.012688000169065265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,6144,0.010808000134097205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,6144,0.007521777517265744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,6144,0.011669333610269757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,5120,0.009654222263230218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,3584,0.007308444215191736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,5120,0.00776800016562144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,5120,0.010751999914646149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,4096,0.008550222549173567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,4096,0.007146666447321574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,4096,0.009483555952707926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,3584,0.008070222205585903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,3584,0.00908799966176351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,3072,0.007522666619883643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,3072,0.006992888947327931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,3072,0.008583111067612966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,2560,0.00685777763525645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,2560,0.006240889016124938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,2560,0.008030222521887885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,2048,0.005551110953092575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,2048,0.007382222347789341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,1536,0.004878222114510006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,1536,0.006934222247865465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,1024,0.004210666649871402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,1024,0.006502222269773483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,768,0.004918222212129169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,768,0.0037840000457233856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,2048,0.00638755535085996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,768,0.006319111006127463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,512,0.0034177777253919174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,512,0.006005333529578314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,256,0.0030817778574095834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,256,0.005736888696750005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,1536,0.005995555470387141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,128,0.00423555572827657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,128,0.0028702221396896574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,1024,0.0052382221652401825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,512,0.0046942221621672315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,256,256,0.004331555631425646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,128,0.005623999983072281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,64,0.002712888850106133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,64,0.005617777920431561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,256,32,0.0027537778433826235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,256,32,0.005640888793600931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,65536,0.019148444135983784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,65536,0.07541600200865003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,16384,0.02289155622323354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,16384,0.009128889275921715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,16384,0.02347555591000451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,12288,0.00812000036239624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,12288,0.019111999207072787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,10240,0.007854222423500484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,10240,0.016767111089494493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,8192,0.013063111239009433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,8192,0.00737955586777793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,8192,0.01365600029627482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,7168,0.011970666547616323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,7168,0.007262222468852997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,7168,0.012522666818565793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,65536,0.07952088779873319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,12288,0.018007111218240526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,6144,0.010755555497275459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,6144,0.006700444552633498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,10240,0.015580443872345818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,6144,0.011439111497667102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,5120,0.007147555549939473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,5120,0.01053155544731352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,4096,0.00673333348499404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,4096,0.009383111364311641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,3584,0.008030222521887885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,3584,0.007104000283612146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,5120,0.00960533320903778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,4096,0.00850755555762185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,3584,0.009026666482289633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,3072,0.007544889218277401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,3072,0.006807111203670502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,3072,0.008465777668688033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,2560,0.006826666494210561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,2560,0.006146666490369373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,2560,0.007884444461928474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,2048,0.0063564446237352155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,2048,0.005479111025730769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,2048,0.007317333585686154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,1536,0.00590133335855272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,1536,0.0047733332547876565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,1536,0.006896889044178857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,1024,0.0052062223354975385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,1024,0.004080888711743885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,1024,0.00645955569214291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,768,0.004886222382386525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,768,0.0037173334923055437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,768,0.006255110932721033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,512,0.004590222405062782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,512,0.0033982222278912864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,512,0.005944889038801193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,256,0.003015111097031169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,256,0.00564177789621883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,128,0.0028444443725877335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,128,0.005497777627574072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,64,0.0026951111439201566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,64,0.005599110904667113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,128,32,0.0026737778551048706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,128,32,0.005538666827811136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,65536,0.016201777590645682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,256,0.004326222257481681
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,65536,0.07516177495320638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,16384,0.008552889029184977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,192,128,128,0.0041928887367248535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,16384,0.02325155503220028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,12288,0.007253333098358578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,12288,0.018586667047606576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,10240,0.007063111497296228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,10240,0.015769778026474845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,8192,0.0069253332912921906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,8192,0.013076444466908773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,7168,0.006783111227883234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,7168,0.012285333540704517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,6144,0.006596444381607904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,6144,0.011260444091426002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,5120,0.0069884442620807225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,5120,0.010443555812040964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,4096,0.006555555595291986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,4096,0.0092604441775216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,3584,0.0070186663005087115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,3584,0.00889066689544254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,3072,0.006712889091836081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,3072,0.008384000096056197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,2560,0.006041777630647023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,2560,0.007823111282454597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,2048,0.005388444496525659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,2048,0.007278222176763747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,1536,0.0047155556579430895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,1536,0.006822222222884496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,1024,0.004042666819360522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,1024,0.006392000036107169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,768,0.003698666476541095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,768,0.006159111029571957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,512,0.0033368888414568375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,512,0.00582755563987626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,256,0.003007111036115222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,256,0.005620444400442972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,128,0.002805333377586471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,128,0.005491555564933353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,64,0.0026897777699761917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,64,0.005512889060709212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,64,32,0.0026568888376156488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,64,32,0.005479111025730769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,65536,0.01534577707449595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,65536,0.07454755571153429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,16384,0.008662222160233391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,16384,0.02314400010638767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,12288,0.007116444408893585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,12288,0.0181040002240075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,10240,0.0069306666652361555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,10240,0.015232000086042615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,8192,0.006800888727108638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,8192,0.01295466638273663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,7168,0.006569777925809224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,7168,0.012132444315486483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,6144,0.006461333483457565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,6144,0.011162666810883416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,5120,0.00693866651919153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,5120,0.010301333334710863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,4096,0.006528000036875407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,4096,0.009146666361225976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,3584,0.006923555499977536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,3584,0.008839999636014303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,3072,0.006701333241330252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,3072,0.008413333031866286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,2560,0.006067555397748947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,2560,0.007824888659848107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,2048,0.005406222409672207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,2048,0.007255111303594377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,1536,0.004709333181381226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,1536,0.006783999916579988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,1024,0.004024888906213972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,1024,0.0063484443558586975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,768,0.003640888879696528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,768,0.00610133343272739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,512,0.0033617777129014335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,512,0.0058168888919883305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,256,0.0029964444951878656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,256,0.005579555614127054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,128,0.002806222273243798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,128,0.005444444302055571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,64,0.002663111107216941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,64,0.005443555613358815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,192,32,32,0.0026826666047175727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,192,32,32,0.005504000104135937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,16384,0.5645439889695909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,16384,0.5278088781568739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,12288,0.41231110360887313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,16384,0.7063057687547473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,12288,0.41994312074449325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,12288,0.5351831118265787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,10240,0.3453973399268256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,10240,0.3573004404703776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,8192,0.2936391035715739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,10240,0.43774843215942383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,8192,0.28558577431572807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,7168,0.24954401122199166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,7168,0.2561448944939507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,8192,0.36481867896185977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,7168,0.3226248953077528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,6144,0.21846755345662436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,6144,0.21847911675771078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,6144,0.2667813301086426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,5120,0.18119555049472383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,5120,0.18277511331770155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,5120,0.22996889220343697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,4096,0.1438257826699151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,4096,0.14875022570292154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,4096,0.18753511375851104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,3584,0.1307448943456014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,3584,0.12894133726755777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,3072,0.11487644248538548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,3584,0.16716978285047743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,3072,0.11177510685390896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,3072,0.14250132772657606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,2560,0.09816622071795994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,2560,0.09211199813418919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,2560,0.12090933322906494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,2048,0.07855644490983751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,2048,0.08149244387944539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,2048,0.09742844767040676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,1536,0.05572444200515747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,1536,0.06430400080151029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,1536,0.07180622551176283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,1024,0.038198222716649376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,1024,0.045787556303872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,1024,0.05144800080193413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,768,0.03170399864514669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,768,0.03585777680079142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,768,0.03970222340689765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,512,0.02442844377623664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,512,0.026910222238964502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,512,0.03382488754060533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,256,0.017139555679427255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,256,0.029496000872717962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,128,0.014833778142929077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,65536,256,0.018956444329685636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,128,0.011644444531864591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,128,0.027786665492587622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,64,0.010004444254769219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,64,0.029576000240114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,65536,32,0.009570666485362584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,65536,32,0.033350222640567355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,65536,0.5786755349900987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,65536,0.7116817898220487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,65536,0.5454782379998101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,16384,0.1452595525317722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,16384,0.17934489250183105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,12288,0.1091466678513421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,12288,0.1087262233098348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,12288,0.13013599978552923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,10240,0.0979200005531311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,10240,0.09001688824759589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,16384,0.14661155806647405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,10240,0.10708889034059312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,8192,0.0744675530327691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,8192,0.07416622506247626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,8192,0.09182488918304443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,7168,0.06625511248906453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,7168,0.07981244723002116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,6144,0.05698044432534111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,6144,0.05579466952217949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,5120,0.049053334527545504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,6144,0.067830224831899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,5120,0.04776444368892246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,5120,0.057914667659335665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,4096,0.039056890540652804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,4096,0.04064977831310696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,4096,0.048483557171291776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,3584,0.03519377774662442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,3584,0.03551466597451104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,7168,0.06509421931372748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,3584,0.04227466715706719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,3072,0.03134844369358487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,3072,0.037831111086739436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,2560,0.026327111654811438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,2560,0.027319111757808264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,3072,0.030392001072565716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,2560,0.03254222207599216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,2048,0.022674666510687932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,2048,0.02250755495495266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,2048,0.02665688925319248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,1536,0.018523555662896898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,1536,0.018241778016090393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,1536,0.021200888686709937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,1024,0.014510222607188754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,1024,0.012964444855848948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,1024,0.016688888271649677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,768,0.010651555326249866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,768,0.015036443869272867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,512,0.011040888726711273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,768,0.012739555703269111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,512,0.008629333641793992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,512,0.013264888690577613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,256,0.006792888873153263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,256,0.010455999937322406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,128,0.007991111112965478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,16384,256,0.009706666900051964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,128,0.004863111095296012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,128,0.009757333331637913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,64,0.004309333446953031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,64,0.010228444304731157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,16384,32,0.004619555754794015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,16384,32,0.011287111375066968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,65536,0.43659199608696836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,65536,0.35503199365403915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,65536,0.5624551243252224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,16384,0.09233155515458848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,16384,0.12414488527509902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,16384,0.13924977514478895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,12288,0.06886933247248332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,12288,0.09706577989790176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,12288,0.10617955525716145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,10240,0.05757244427998861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,10240,0.08783466948403253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,10240,0.08643733130560981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,8192,0.07152444124221802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,8192,0.0705146657096015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,7168,0.04176888863245646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,8192,0.04807644420199924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,7168,0.05425955520735847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,7168,0.06259377797444661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,6144,0.04667644368277656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,6144,0.05423644516203138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,5120,0.03954755597644382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,6144,0.035749332772360906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,5120,0.031033777528338965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,5120,0.04642400145530701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,4096,0.03255733185344272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,4096,0.039453334278530545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,3584,0.023634667197863262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,3584,0.029275556405385334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,3584,0.03581777877277798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,4096,0.025898666845427618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,3072,0.025916443930731878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,3072,0.030750221676296655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,3072,0.021606221795082092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,2560,0.019440889358520508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,2560,0.021183111601405676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,2560,0.026366222235891554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,2048,0.016920000314712524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,2048,0.017649778061442904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,2048,0.023028444912698533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,1536,0.014582221706708273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,1536,0.013891556196742587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,1536,0.01850044396188524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,1024,0.01034577770365609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,1024,0.014881778094503613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,768,0.011010666688283285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,768,0.00885333369175593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,768,0.013510222236315409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,512,0.010104888843165504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,512,0.007350222104125553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,256,0.005335111171007156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,512,0.011375111010339526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,256,0.008108444511890411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,256,0.009518221848540837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,1024,0.0122417774465349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,12288,128,0.0064275554484791225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,128,0.00434933313065105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,128,0.009004444711738164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,64,0.004156444635656145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,64,0.009296889106432596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,12288,32,0.004274666723277834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,12288,32,0.00998577806684706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,65536,0.32882576518588597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,65536,0.38938577969868976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,65536,0.4698746469285753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,16384,0.08639999892976548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,16384,0.1082657774289449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,16384,0.11855555905236138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,12288,0.0653582215309143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,12288,0.08713244729571873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,12288,0.08672266536288792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,10240,0.05415999889373779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,10240,0.07612266805436876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,10240,0.07066222031911214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,8192,0.045221332046720714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,8192,0.06255200174119738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,8192,0.058000889089372426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,7168,0.03929333223236932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,7168,0.05675111214319865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,7168,0.0521031121412913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,6144,0.03363110952907138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,6144,0.04714577727847629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,6144,0.045610666275024414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,5120,0.040417777167426214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,5120,0.03866844375928243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,4096,0.024215110474162634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,4096,0.03351199958059523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,4096,0.032391111056009926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,3584,0.02238133384121789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,3584,0.029765334394243028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,3584,0.03010400136311849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,3072,0.020224889119466145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,3072,0.026047999660174053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,3072,0.027456000447273254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,2560,0.017807111144065857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,5120,0.029113776153988306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,2560,0.01901155544651879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,2560,0.022595556245909795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,2048,0.015467555986510383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,2048,0.01591911084122128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,2048,0.019197333190176222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,1536,0.013444444371594323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,1536,0.012699555191728802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,1536,0.0163111107216941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,1024,0.011152000062995486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,1024,0.009940444595283931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,1024,0.013376889129479727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,768,0.008585777547624376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,768,0.012410666379663678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,768,0.010139555566840703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,512,0.009328000247478485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,512,0.007345777418878343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,512,0.009956444303194681
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,256,0.007117333511511485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,256,0.005098666581842634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,256,0.00867911097076204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,10240,128,0.005992888990375731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,128,0.004269333349333869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,128,0.008276444342401292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,64,0.004266666869322459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,64,0.00847111145655314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,10240,32,0.003926222109132343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,10240,32,0.008966222405433655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,65536,0.2281306584676107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,65536,0.2953902350531684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,65536,0.36392977502610946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,16384,0.059241778320736356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,16384,0.07751466830571492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,16384,0.09132177962197198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,12288,0.044884443283081055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,12288,0.056715554661220975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,12288,0.06765866941875882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,10240,0.03875911235809326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,10240,0.049125333627065025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,10240,0.05476977758937412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,8192,0.032009777095582746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,8192,0.03872266742918227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,8192,0.046741333272722035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,7168,0.028902222712834675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,7168,0.0340008901225196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,7168,0.042212443219290845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,6144,0.02551377813021342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,6144,0.03014844324853685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,6144,0.03558399942186143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,5120,0.02604177759753333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,5120,0.022544888986481562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,5120,0.0314337776766883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,4096,0.019336000084877014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,4096,0.02219555609756046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,4096,0.026234666506449383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,3584,0.01807466646035512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,3584,0.019426667028003268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,3584,0.02350577712059021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,3072,0.016340444485346477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,3072,0.017229333519935608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,3072,0.021433777279324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,2560,0.014953777194023132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,2560,0.015053333507643806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,2560,0.018723555737071566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,2048,0.013440888788965015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,2048,0.012396444049146442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,2048,0.01606399979856279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,1536,0.011870221959220039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,1536,0.010226666927337646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,1536,0.013713777893119387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,1024,0.010239111052619087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,1024,0.00812977800766627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,1024,0.01182666669289271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,768,0.007152000235186682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,768,0.010499555203649731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,512,0.007436444362004598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,512,0.006055111272467508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,512,0.008691555923885768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,256,0.0058453331391016645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,768,0.009148444566461775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,256,0.004322666674852371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,256,0.007827555967701806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,128,0.003839111162556542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,128,0.007446222007274628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,64,0.003539555602603488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,64,0.00775999989774492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,8192,32,0.003930666794379552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,8192,32,0.008328888979223039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,65536,0.2606408860948351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,65536,0.3583208984798855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,8192,128,0.00534400012758043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,65536,0.2980915440453423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,16384,0.08058489031261869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,16384,0.07635821898778279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,16384,0.08560978041754828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,12288,0.05884444713592529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,12288,0.05089777708053589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,12288,0.0652373300658332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,10240,0.04939377639028761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,10240,0.04464266697565714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,10240,0.053605334626303784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,8192,0.03532444437344869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,8192,0.04478933413823446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,8192,0.04180533356136746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,7168,0.03492000036769443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,7168,0.03160088923242357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,7168,0.041125333971447416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,6144,0.02920266654756334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,6144,0.02769600020514594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,6144,0.03356800145573086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,5120,0.025778666138648987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,5120,0.024304888314670984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,5120,0.031090666850407917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,4096,0.02008800043000115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,4096,0.025250666671329077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,4096,0.02236977716286977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,3584,0.01960000064637926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,3584,0.018061333232455783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,3584,0.02373955481582218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,3072,0.017345777816242643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,3072,0.015999999311235216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,3072,0.020111999577946134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,2560,0.015306666493415833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,2560,0.013707555830478668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,2560,0.018212444252438016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,2048,0.013359111216333179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,2048,0.01165066659450531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,2048,0.015396444333924187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,1536,0.009733333355850643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,1536,0.013051555388503604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,1024,0.010268444816271463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,1024,0.007882666256692674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,1024,0.011586666935020022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,768,0.0069919998447100324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,768,0.009906666974226633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,512,0.007517333659860823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,512,0.00573955559068256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,512,0.008559111091825698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,1536,0.011728888584507836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,768,0.008777778181764815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,256,0.006095110956165526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,256,0.004300444490379757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,256,0.007803555164072249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,7168,128,0.005635555419656966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,128,0.0037804444630940757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,128,0.007354666789372762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,64,0.0034657776769664553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,64,0.007627555893527136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,7168,32,0.003560888684458203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,7168,32,0.008015111088752747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,65536,0.18467733595106336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,65536,0.2247671021355523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,65536,0.34676088227166074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,16384,0.04977066649330986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,16384,0.0676471127404107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,16384,0.0843413339720832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,12288,0.03923733366860284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,12288,0.05475022064314949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,12288,0.06300710969501071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,10240,0.03975377811325921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,10240,0.050274666812684804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,8192,0.03316266669167413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,10240,0.033732444047927856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,8192,0.028147555059856836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,8192,0.04263822237650553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,7168,0.025634666283925373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,7168,0.029491557015313044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,7168,0.03802666730350918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,6144,0.022566222482257422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,6144,0.026122666067547266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,6144,0.03277511066860623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,5120,0.02015733387735155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,5120,0.02238222294383579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,5120,0.028596444262398615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,4096,0.01884888940387302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,4096,0.017472000585661996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,4096,0.023403555154800415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,3584,0.01679733395576477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,3584,0.0215013325214386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,3072,0.014999111493428549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,3072,0.014811555544535318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,3072,0.019719999697473314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,2560,0.013793778088357715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,3584,0.015987555185953777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,2560,0.012864000267452665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,2560,0.017293334007263184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,2048,0.012359110845459832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,2048,0.01110666659143236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,2048,0.014544889330863953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,1536,0.00944622192117903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,1536,0.01090222183201048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,1536,0.012548444171746572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,1024,0.01000711073478063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,1024,0.007671111159854465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,1024,0.010508444574144153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,768,0.006909333169460297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,768,0.009256000320116678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,768,0.007803555164072249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,512,0.006471111128727595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,512,0.005606222069925732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,512,0.008444444172912175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,256,0.005592888842026393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,256,0.004260444392760594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,256,0.007744889292452071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,6144,128,0.005146666533417172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,128,0.003801777958869934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,128,0.00741511086622874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,64,0.0034506666577524612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,64,0.0075351107451650835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,6144,32,0.0035342222286595237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,6144,32,0.007696000238259633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,16384,0.05680177609125773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,16384,0.0732311142815484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,65536,0.19240800539652506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,65536,0.24637156062655977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,65536,0.26843200789557564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,16384,0.061944888697730176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,12288,0.05368177758322822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,12288,0.045647111203935414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,12288,0.04708355665206909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,10240,0.04347199863857693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,10240,0.043134222428003945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,10240,0.038888888226615057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,8192,0.037406222687827215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,8192,0.03498044278886583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,8192,0.03202577763133579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,7168,0.03237333231502109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,7168,0.03250311149491204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,7168,0.03001333276430766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,6144,0.028936889436509874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,6144,0.024832000335057575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,5120,0.023168888356950548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,6144,0.027514666318893433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,5120,0.02104355560408698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,5120,0.022708444131745234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,4096,0.019754666421148512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,4096,0.017852443787786696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,4096,0.01907111042075687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,3584,0.018094221750895183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,3584,0.015607999430762397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,3584,0.017704889178276062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,3072,0.016297777493794758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,3072,0.016677333248986136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,3072,0.016167999969588388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,2560,0.013878222141000958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,2560,0.01257155587275823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,2560,0.014156444205178155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,2048,0.01088355564408832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,2048,0.012362666428089142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,1536,0.010325333310498131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,1536,0.00923111124171151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,2048,0.012097777591811286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,1536,0.011286222272449069
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,1024,0.008489777644475302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,1024,0.007589333587222629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,1024,0.008935111264387766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,768,0.006559111177921295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,768,0.008265777594513362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,512,0.0064275554484791225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,512,0.005516444643338521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,768,0.007311999797821045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,512,0.0073724447025193115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,256,0.005718222094906702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,256,0.004364444563786189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,256,0.0068959999415609576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,5120,128,0.005220444252093633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,128,0.003916444463862313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,128,0.006638222270541721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,64,0.0034533333447244433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,64,0.0069013333155049225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,5120,32,0.003548444559176763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,5120,32,0.007012444237867991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,65536,0.1510044468773736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,16384,0.04297777679231432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,65536,0.15184356106652155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,65536,0.23738222652011445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,16384,0.045383999745051064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,16384,0.05919733312394884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,12288,0.03342311249838935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,12288,0.03624355461862352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,12288,0.0417520006497701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,10240,0.028836445675955877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,10240,0.03230933348337809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,10240,0.035511109564039446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,8192,0.023894222246276006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,8192,0.028163555595609877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,8192,0.031015998787350122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,7168,0.021645334031846788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,7168,0.025407110651334126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,7168,0.027653333213594224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,6144,0.019347555107540555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,6144,0.023267555567953322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,6144,0.023543111152119104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,5120,0.017234666479958426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,5120,0.020908445119857788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,5120,0.022169777088695105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,4096,0.014997333288192749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,4096,0.01974311139848497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,4096,0.018046221799320646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,3584,0.013883555928866068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,3584,0.017121778594122995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,3584,0.016559999850061204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,3072,0.012693333129088083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,3072,0.015166223049163818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,3072,0.014781332678265043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,2560,0.01183377785815133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,2560,0.010562666588359408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,2560,0.013305777476893531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,2048,0.010486221975750394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,2048,0.008960888617568547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,2048,0.012002666791280111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,1536,0.0092275556590822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,1536,0.007819555699825287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,1536,0.010743111371994019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,1024,0.006151999864313338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,1024,0.006928000185224745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,1024,0.00869955536392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,768,0.006145777801672618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,768,0.005160000175237656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,768,0.007991111112965478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,512,0.004501333253251182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,512,0.007326222128338284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,256,0.003882666842805015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,256,0.00692622239391009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,128,0.0048177776237328844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,512,0.005595555735958948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,128,0.0034799998005231223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,128,0.006601777755551868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,4096,256,0.005144000053405762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,64,0.0033048888047536216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,64,0.006771555377377405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,4096,32,0.00350222239891688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,4096,32,0.006822222222884496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,65536,0.14801600244310167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,65536,0.14602933989630804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,16384,0.04165600074662103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,65536,0.22947110070122612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,16384,0.04093955622778998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,16384,0.057977775732676186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,12288,0.03253511256641812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,12288,0.03360889024204678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,12288,0.04266666703754001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,10240,0.030047112041049536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,10240,0.034824000464545354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,8192,0.023149333066410486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,10240,0.027668444646729365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,8192,0.025960000024901494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,8192,0.029456890291637842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,7168,0.021015110943052504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,7168,0.023603555228975084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,7168,0.026123555170165166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,6144,0.018969777557584975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,6144,0.021524444222450256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,6144,0.024302222662501868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,5120,0.016878222425778706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,5120,0.01924444403913286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,5120,0.020806221498383414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,4096,0.012535110943847232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,4096,0.01771377854877048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,3584,0.013642667068375481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,3584,0.011262222296661802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,3584,0.016327111257447135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,4096,0.014579556054539151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,3072,0.01017155581050449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,3072,0.01455200049612257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,2560,0.011605333122942181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,2560,0.009131555755933126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,2560,0.013259555730554791
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,3072,0.012479999827014076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,2048,0.010428444378905825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,2048,0.008051555189821456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,2048,0.0118568887313207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,1536,0.008600888980759515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,1536,0.00702311098575592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,1536,0.009673777553770278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,1024,0.006583111153708563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,1024,0.0058311112225055695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,1024,0.00851733320289188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,768,0.005939555664857228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,768,0.0047591109242704184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,768,0.007984889050324758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,512,0.005519999812046687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,512,0.004085333396991094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,512,0.0074026667409473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,64,0.003104000041882197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,256,0.005053333524200651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,256,0.003579555700222651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,256,0.006791111081838608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,128,0.0033368888414568375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,128,0.006523555351628198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3584,128,0.004755555755562252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,64,0.0067528887755341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3584,32,0.0032195556494924757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3584,32,0.006908444480763541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,65536,0.23028000195821127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,65536,0.12028355068630642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,16384,0.03928000066015456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,65536,0.1820248895221286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,16384,0.049025777313444346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,16384,0.06163022253248426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,12288,0.04670222269164192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,12288,0.03103733393881056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,12288,0.037010666396882795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,10240,0.03912711143493652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,10240,0.027644443843099806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,10240,0.031312889522976346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,8192,0.03256888853179084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,8192,0.02401155564520094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,8192,0.02698488864633772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,7168,0.02862311071819729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,7168,0.021778666310840186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,7168,0.02385422256257799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,6144,0.02442844377623664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,6144,0.019886222150590684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,6144,0.02131555477778117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,5120,0.020799999435742695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,5120,0.01774044500456916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,5120,0.01901955571439531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,4096,0.01796711153454251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,4096,0.016554666890038382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,4096,0.01646933290693495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,3584,0.016395555602179635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,3584,0.014239999983045789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,3584,0.015153777268197803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,3072,0.012991111311647626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,3072,0.013957333233621387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,3072,0.014532445205582513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,2560,0.00889244427283605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,2560,0.01272266689274046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,2560,0.012929778132173749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,2048,0.011209777659840055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,2048,0.00795555528667238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,2048,0.010818666881985135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,1536,0.00846400029129452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,1536,0.006968888971540663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,1536,0.009218666288587783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,1024,0.007070221834712558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,1024,0.0052897776994440295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,1024,0.00835111074977451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,768,0.006451555424266391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,768,0.004588444613748127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,768,0.00776800016562144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,512,0.00590133335855272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,512,0.0040693332751592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,512,0.007184888753626082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,256,0.0053004444473319584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,256,0.003583111282851961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,256,0.00675999994079272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,3072,128,0.004972444640265571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,128,0.0032968889507982465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,128,0.00647199981742435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,64,0.003117333269781537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,64,0.006647999915811751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,3072,32,0.0031831111345026228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,3072,32,0.006744888921578725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,16384,0.03439911206563314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,65536,0.10188888841205174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,65536,0.12556089295281306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,65536,0.12418488661448161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,16384,0.03184355629814996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,16384,0.03218844532966614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,12288,0.025253333979182776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,12288,0.02646844420168135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,12288,0.02658933401107788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,10240,0.019945777124828763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,10240,0.023522666758961145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,10240,0.023129777775870428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,8192,0.01885688967174954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,8192,0.020375111036830477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,8192,0.019890666007995605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,7168,0.018611555298169453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,7168,0.01811377704143524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,6144,0.014132445057233175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,6144,0.01658933361371358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,6144,0.0163128889269299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,7168,0.015663999650213454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,5120,0.01571644511487749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,5120,0.014555555250909595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,4096,0.013942221800486246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,4096,0.012709333664841123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,3584,0.011919111013412476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,3584,0.01200799975130293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,3072,0.010955555571450127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,3072,0.011152889165613385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,2560,0.010204444328943888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,5120,0.012812444733248817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,4096,0.011483555866612328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,3584,0.01096088853147295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,2560,0.009194666312800514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,3072,0.010096000300513374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,2560,0.009687999884287516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,2048,0.008016000191370646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,2048,0.008336000144481659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,2048,0.008771555291281806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,1536,0.006955555743641323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,1536,0.007152000235186682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,1536,0.007806222471925948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,1024,0.005969777703285217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,1024,0.005894222193294101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,1024,0.007099555598364935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,768,0.005052444421582752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,768,0.006837333242098491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,512,0.004392000122202767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,512,0.0052915554907586836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,512,0.0064382221963670515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,256,0.00490133340160052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,256,0.0036755556033717263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,256,0.006112888869312074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,768,0.00556355549229516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2560,128,0.00463733325401942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,128,0.0033440000067154565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,128,0.005816000203291575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,64,0.0030817778574095834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,64,0.00601333338353369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2560,32,0.003136000078585413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2560,32,0.006140444427728653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,65536,0.12281866868336995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,65536,0.08293155829111735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,16384,0.030807998445298936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,65536,0.10482666889826457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,16384,0.02717777755525377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,16384,0.030233777231640283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,12288,0.023453333311610754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,12288,0.02219999995496538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,12288,0.022859555151727464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,10240,0.01973511113060845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,10240,0.01953155630164676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,10240,0.01959377858373854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,8192,0.017335999343130324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,8192,0.016892444756295946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,8192,0.01808088852299584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,7168,0.015574221809705099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,7168,0.016138666205936007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,6144,0.014322666658295525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,6144,0.014744889405038623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,5120,0.012319111161761813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,5120,0.013599110974205865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,7168,0.015483554866578845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,5120,0.013380444712109037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,4096,0.011019555230935415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,6144,0.013635555903116861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,4096,0.012120000190205045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,4096,0.011752000285519494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,3584,0.010761777559916178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,3584,0.011262222296661802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,3072,0.009518221848540837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,3072,0.010114666488435533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,2560,0.00870400004916721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,2560,0.008251555263996124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,3584,0.010514666636784872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,3072,0.009693333672152625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,2560,0.00979644474056032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,2048,0.007445333732499017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,2048,0.007019555403126611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,2048,0.008119111259778341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,1536,0.006631111105283101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,1536,0.0063528890411059065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,1536,0.007481777833567725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,1024,0.0058311112225055695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,1024,0.00463733325401942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,1024,0.0069413334131240845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,768,0.005499555418888728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,768,0.004311999926964442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,768,0.006709333509206772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,512,0.003866666720973121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,512,0.006320888797442119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,256,0.004853333450025982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,256,0.0034844444857703317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,256,0.005951111101441913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,512,0.005281777845488654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,2048,128,0.004603555632962121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,128,0.0031911111954185697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,128,0.005757333503829108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,64,0.002995555599530538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,64,0.005920889063013925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,2048,32,0.003185777821474605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,2048,32,0.005984000033802456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,65536,0.09319288863076104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,65536,0.06928355826271905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,65536,0.08288800054126315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,16384,0.023224888576401606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,16384,0.02548888822396596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,16384,0.026915555198987324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,12288,0.01849244369400872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,12288,0.02033155494266086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,10240,0.019052444232834708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,10240,0.01611111064751943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,10240,0.01805688937505086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,12288,0.021752888957659405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,8192,0.01424088908566369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,8192,0.015825778245925903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,8192,0.015968888998031616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,7168,0.013721778160995908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,7168,0.014882667197121514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,7168,0.014840889308187695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,6144,0.013467555244763693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,6144,0.015136000182893542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,6144,0.013794666363133324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,5120,0.01222222215599484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,5120,0.01331911070479287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,5120,0.012823111481136747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,4096,0.01069777790043089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,4096,0.01148000028398302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,3072,0.009569777382744683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,4096,0.011228444675604502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,3584,0.01053600013256073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,3584,0.010175111393133799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,3072,0.009480888644854227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,2560,0.007921777665615082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,2560,0.007734222544564142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,2560,0.008661333057615492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,3584,0.010114666488435533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,3072,0.008775110873911116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,2048,0.007267555428875818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,2048,0.006821333534187741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,2048,0.007805333369308048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,1536,0.005399999933110342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,1536,0.007353777686754863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,1536,0.006534222099516127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,1024,0.005800888770156437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,1024,0.004543111142185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,1024,0.006800888727108638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,768,0.005398222141795688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,768,0.004221333397759331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,768,0.006600889066855113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,512,0.005048888838953442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,512,0.0037955554823080697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,512,0.006228444476922353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,256,0.0034026667061779234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,256,0.005903111149867375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,128,0.0031031111462248694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,128,0.005689777847793367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,64,0.0029351111087534162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,64,0.00591822216908137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1536,32,0.003000000077817175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1536,32,0.0058755555914507965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,65536,0.18475288814968535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,256,0.0047360000511010485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,65536,0.04696177774005466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,65536,0.07625866598553128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,16384,0.01753155555990007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,16384,0.023740443918440077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,12288,0.03742311067051358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,12288,0.014388443695174323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,12288,0.019448889626397025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,10240,0.031403554810418025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1536,128,0.004521777646409141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,10240,0.01382844481203291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,10240,0.017599110802014668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,8192,0.01331555595000585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,16384,0.04960088928540548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,8192,0.015327110886573792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,7168,0.023581332630581323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,7168,0.013008889224794177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,7168,0.0143857780430052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,6144,0.020566221740510728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,6144,0.012264000044928657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,6144,0.013295110729005603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,5120,0.011165333290894827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,8192,0.025851555996470984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,5120,0.012206222448084088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,4096,0.014194667339324951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,4096,0.00982844415638182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,4096,0.010282666318946415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,5120,0.01715555621518029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,3584,0.012854222622182635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,3584,0.00906844437122345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,3584,0.009535999761687385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,3072,0.01143555591503779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,3072,0.00831911133395301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,3072,0.008859555754396651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,2560,0.01015022231472863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,2560,0.0074648890230390764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,2560,0.008254222571849823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,2048,0.0063484443558586975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,2048,0.007673777639865875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,1536,0.007817777494589487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,1536,0.0052826665341854095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,1536,0.007194666398896112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,1024,0.006505777852402792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,2048,0.008973333570692275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,1024,0.004493333399295807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,1024,0.006673777682913675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,768,0.0058764442801475525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,768,0.0041617775956789655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,768,0.006485333459244833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,512,0.003742222156789568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,512,0.006158222340875202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,512,0.005430222385459476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,256,0.004953777624501122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,256,0.0032879999942249725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,256,0.0058755555914507965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,128,0.0030542222989930045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,128,0.005702222386995952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,64,0.00286666676402092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,1024,128,0.004701333327425851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,64,0.00573955559068256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,1024,32,0.00291377781993813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,1024,32,0.005713777823580637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,65536,0.15139555931091309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,65536,0.0396382212638855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,65536,0.07490844196743436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,16384,0.04472088813781738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,16384,0.014917333920796713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,16384,0.02365866634580824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,12288,0.03461688756942749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,12288,0.015087111128701104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,12288,0.01945688823858897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,10240,0.015025777949227227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,10240,0.017652443713612027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,8192,0.025014221668243408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,8192,0.013097777962684631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,10240,0.029250668154822454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,8192,0.01517688896920946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,7168,0.022348443667093914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,7168,0.011737777955002256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,7168,0.014198222094111972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,6144,0.011160888605647616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,6144,0.013095999757448832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,6144,0.018703111343913607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,5120,0.016372443901167977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,5120,0.010469333165221743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,5120,0.011760888828171624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,4096,0.013565333353148567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,4096,0.009563555320103964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,4096,0.009970666633711921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,3584,0.012443555725945367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,3584,0.009278222090668148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,3584,0.009509333305888707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,3072,0.010868444210953183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,3072,0.008623110751310984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,3072,0.008770667016506195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,2560,0.009376889301670922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,2560,0.007141333487298753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,2560,0.008272888759771982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,2048,0.00830666638082928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,2048,0.005818666683302985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,2048,0.007673777639865875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,1536,0.007254222200976477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,1536,0.005103111267089844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,1536,0.007143110864692264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,1024,0.006136000156402588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,1024,0.00432977784011099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,1024,0.006683555742104848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,768,0.005678222411208683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,768,0.00397866674595409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,768,0.006485333459244833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,512,0.00360000009338061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,512,0.005400889035728242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,512,0.006105777704053455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,256,0.004827555682924059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,256,0.003232888877391815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,256,0.0058319999112023254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,128,0.0029528888149393927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,768,128,0.004583111239804162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,128,0.005671111245950063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,64,0.002798222212327851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,64,0.005706666658322017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,768,32,0.0028577778074476454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,768,32,0.005695999910434087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,65536,0.0780844423505995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,65536,0.029574222034878198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,65536,0.07514311207665338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,16384,0.022924444741672937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,16384,0.014160000615649752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,16384,0.023342221975326538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,12288,0.018631999691327412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,12288,0.012056888805495368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,12288,0.019259555472267997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,10240,0.016389333539538913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,10240,0.01110755569405026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,10240,0.017320000463061862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,8192,0.010478222535716163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,8192,0.014911111858155994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,8192,0.013610666824711693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,7168,0.012443555725945367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,7168,0.010233778092596266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,7168,0.013712888790501488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,6144,0.010891555911964841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,6144,0.009472000102202097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,6144,0.012373333175977072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,5120,0.008919999831252629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,5120,0.009703999592198266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,5120,0.010969777901967367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,4096,0.008711111214425828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,4096,0.007998222278224098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,4096,0.009641778137948778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,3584,0.008247999681366814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,3584,0.008099555969238281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,3584,0.009191110730171204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,3072,0.007582222421964009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,3072,0.007290666302045186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,3072,0.008745777938101027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,2560,0.007034666836261749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,2560,0.006351111249791251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,2560,0.008045333127180735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,2048,0.006533333410819371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,2048,0.005650666852792104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,2048,0.0075093333919843035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,1536,0.006039999839332368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,1536,0.004955555415815777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,1536,0.007055111229419708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,1024,0.005422222117582957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,1024,0.0042257776690853965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,1024,0.006559999866618051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,768,0.005065777649482091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,768,0.003903111235962974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,768,0.006342222293217977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,512,0.0035048888789282907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,512,0.004820444517665439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,512,0.006034666465388403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,256,0.004523555437723796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,256,0.0031271111220121384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,256,0.005763555566469829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,128,0.0028693332440323303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,128,0.005608888963858287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,64,0.0027742222365405825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,512,128,0.004292444636424382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,64,0.005630222045713001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,512,32,0.0027884443600972495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,512,32,0.005646222167544895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,65536,0.07692799965540568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,65536,0.025311110748185053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,65536,0.07476177480485705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,16384,0.01111822244193819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,16384,0.0232933329211341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,12288,0.018195556269751657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,12288,0.009929777847396003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,12288,0.019088000059127808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,16384,0.022666666242811415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,10240,0.009287999735938178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,10240,0.017122666041056316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,10240,0.015562666787041558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,8192,0.012949333422713809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,8192,0.008399110701349046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,8192,0.0136808885468377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,7168,0.011716444459226398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,7168,0.007823111282454597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,7168,0.012777778009573618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,6144,0.0074746666683091064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,6144,0.011607999602953592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,5120,0.009538667069541084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,5120,0.007680888805124495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,5120,0.010811555716726514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,6144,0.010567111273606619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,4096,0.00850666645500395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,4096,0.007254222200976477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,4096,0.009664889011118147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,3584,0.008094222181373173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,3584,0.007272000114123027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,3584,0.008986666798591614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,3072,0.007382222347789341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,3072,0.006897777732875612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,3072,0.008787555827034844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,2560,0.006853333363930385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,2560,0.006250666661394968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,2560,0.008055999875068665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,2048,0.005572444448868434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,2048,0.007445333732499017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,1536,0.005752000129885144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,1536,0.004907555464241239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,1536,0.0069466663731469055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,2048,0.0063697778516345555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,1024,0.005244444641802046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,1024,0.0041493334703975255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,1024,0.0064222220745351576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,768,0.004946666873163647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,768,0.0037928890022966596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,768,0.006324444380071428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,512,0.004658666749795278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,512,0.0034417777011791864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,512,0.005988444305128521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,256,0.0030737777964936364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,256,0.005713777823580637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,128,0.004200889004601372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,128,0.0028764444092909494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,128,0.00563733321097162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,64,0.0026835555003749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,64,0.005532444351249271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,256,32,0.002740444408522712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,256,256,0.004392000122202767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,256,32,0.005657777604129579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,65536,0.01817599932352702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,65536,0.07450311051474677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,16384,0.022414222359657288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,65536,0.07702844672732882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,16384,0.009720000127951304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,16384,0.023190221852726404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,12288,0.017616000440385606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,12288,0.0079644446571668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,12288,0.018412444326612685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,10240,0.015171556009186639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,10240,0.008046222229798635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,10240,0.015648888217078317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,8192,0.012694222231705984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,8192,0.007341333561473423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,8192,0.013453333742088742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,7168,0.01165511127975252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,7168,0.007093333535724216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,7168,0.01239733315176434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,6144,0.006716444260544247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,6144,0.01145511120557785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,5120,0.009496000077989366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,5120,0.007117333511511485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,5120,0.010565333068370819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,4096,0.008451555338170793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,4096,0.006746666712893381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,4096,0.00941244430012173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,6144,0.01052266690466139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,3584,0.007150222029950883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,3584,0.00903555585278405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,3584,0.007930667036109501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,3072,0.007354666789372762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,3072,0.006871999965773688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,3072,0.008480889101823172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,2560,0.006154666758245892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,2560,0.007869333028793335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,2048,0.006271111054552927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,2048,0.005511999958091312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,2048,0.007227555745177799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,1536,0.00573955559068256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,1536,0.004771555463473002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,1536,0.006872889068391588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,1024,0.00508977762526936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,1024,0.004053333153327306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,1024,0.0064444442590077715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,768,0.003719999972316954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,2560,0.006772444479995304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,768,0.006187555690606435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,512,0.0046160001721647055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,768,0.004899555610285865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,512,0.0033857776886887024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,512,0.0058808889653947614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,256,0.0030737777964936364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,256,0.005618666609128316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,128,0.0028275555620590844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,128,0.005477333234416113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,64,0.0026897777699761917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,64,0.005514666852023866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,128,32,0.002708444371819496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,128,32,0.005491555564933353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,65536,0.015291555060280694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,65536,0.07451377974616157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,256,0.004316444612211651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,160,128,128,0.004159111115667555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,16384,0.008082666330867344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,16384,0.022664889693260193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,12288,0.007027555671003129
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,12288,0.017700443665186565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,10240,0.007000000112586551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,10240,0.01499022212293413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,8192,0.0069439998931354946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,8192,0.01298577752378252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,7168,0.006788444601827198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,7168,0.012159999873903064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,6144,0.006573333508438534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,6144,0.011147555377748279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,5120,0.0070257774657673305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,5120,0.010296000374688042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,4096,0.006591111007663939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,4096,0.009176000124878353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,3584,0.006986666884687211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,3584,0.008796444369686974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,3072,0.006686222222116258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,3072,0.008319999608728621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,2560,0.006059555543793573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,2560,0.007748444047239091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,2048,0.005385777602593104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,2048,0.007269333634111617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,1536,0.004726222405831019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,1536,0.006821333534187741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,1024,0.004027555386225383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,1024,0.006275555739800136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,768,0.003660444584157732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,768,0.006157333238257303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,512,0.0033288889875014624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,512,0.0058497778243488735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,256,0.002990222225586573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,256,0.005585777676767773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,128,0.002806222273243798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,128,0.005424889011515512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,64,0.0026480000880029467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,64,0.005506666584147348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,64,32,0.0026497778793176017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,64,32,0.005459555735190709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,65536,0.014771555860837301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,65536,0.07407199674182467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,16384,0.0079644446571668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,16384,0.022827555735905964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,12288,0.007312888900438945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,12288,0.017131555411550734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,10240,0.0069439998931354946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,10240,0.015180443723996481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,8192,0.006687110910813014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,8192,0.012989333106411828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,7168,0.006609777609507243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,7168,0.01218311074707243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,6144,0.006423111177153057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,6144,0.011147555377748279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,5120,0.0069208890199661255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,5120,0.010421333213647207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,4096,0.00639111093348927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,4096,0.00921688891119427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,3584,0.006911110960774952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,3584,0.008750222623348236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,3072,0.00665866666369968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,3072,0.008273777862389883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,2560,0.006080889039569431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,2560,0.007840000092983246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,2048,0.005382222019963794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,2048,0.0072266666425599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,1536,0.004707555390066571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,1536,0.006802666518423293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,1024,0.004025777594910728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,1024,0.006315555423498154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,768,0.003626666549179289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,768,0.006150222072998683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,512,0.003351111171974076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,512,0.0058053334554036455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,256,0.0030168888883458245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,256,0.005572444448868434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,128,0.002805333377586471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,128,0.005427555491526921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,64,0.002630222174856398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,64,0.005505777895450592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,160,32,32,0.00263288886182838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,160,32,32,0.005425777700212266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,16384,0.5397129058837891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,16384,0.3539884355333116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,12288,0.40721866819593644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,16384,0.27683912383185494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,12288,0.20936889118618437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,12288,0.25372444258795845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,10240,0.3425760004255507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,10240,0.2075288825564914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,8192,0.274183988571167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,8192,0.1714248922136095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,7168,0.13352889484829372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,7168,0.24448355038960776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,7168,0.14640266365475124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,6144,0.21161333719889322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,6144,0.11300888326432969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,10240,0.17882311344146729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,6144,0.12405867046780056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,8192,0.14664888381958008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,5120,0.10102044211493598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,5120,0.17600088649325898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,4096,0.14338844352298313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,5120,0.10363733106189305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,4096,0.07970133092668322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,4096,0.08436622222264607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,3584,0.0716186695628696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,3584,0.1263315545188056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,3584,0.0748586654663086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,3072,0.11016711261537339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,3072,0.06680799855126275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,3072,0.06703199942906697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,2560,0.05488444368044535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,2560,0.09329244163301255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,2560,0.05589955382876926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,2048,0.04554933309555054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,2048,0.07640355825424194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,2048,0.04765599966049194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,1536,0.037032888995276556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,1536,0.06318133407168917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,1536,0.038570665650897555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,1024,0.02827022141880459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,1024,0.04300355580117968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,1024,0.030879110097885132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,768,0.025099555651346844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,768,0.0362044440375434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,768,0.028220444917678833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,512,0.02516977820131514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,512,0.025342222717073228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,512,0.02017511096265581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,256,0.015418666932317944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,256,0.015916445189052157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,256,0.022375999225510493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,65536,128,0.01239111108912362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,128,0.010446222292052375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,32,0.008410666551854875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,128,0.020989333589871723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,65536,64,0.008861333131790161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,64,0.02330488959948222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,65536,32,0.024140444066789415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,65536,0.557436466217041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,65536,0.29083289040459526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,65536,0.38501866658528644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,16384,0.07665333482954237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,16384,0.1347715589735243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,16384,0.09460000197092693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,12288,0.057966219054328076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,12288,0.10263644324408637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,12288,0.06844800048404269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,10240,0.04981866810056898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,10240,0.09672000010808308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,10240,0.05563288927078247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,8192,0.07124800152248807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,8192,0.04187466700871786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,8192,0.04723555511898465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,7168,0.03702044486999512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,7168,0.06599733564588758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,7168,0.04039644532733493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,6144,0.05428444345792135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,6144,0.035622222556008235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,5120,0.028817776176664565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,5120,0.04600711001290215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,6144,0.032905777295430504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,5120,0.031116445859273274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,4096,0.024509333901935156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,4096,0.0384115543630388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,4096,0.025957332717047796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,3584,0.02251911163330078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,3584,0.033964445193608604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,3584,0.023914666639433965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,3072,0.020435555113686454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,3072,0.029847999413808186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,3072,0.021337777376174927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,2560,0.01835555500454373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,2560,0.026208889153268602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,2560,0.019221333993805777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,2048,0.015992888145976596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,2048,0.021801778011851843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,2048,0.01692800058258904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,1536,0.0177520000272327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,1536,0.0143857780430052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,1024,0.011557333171367645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,1024,0.012548444171746572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,1536,0.013923555612564087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,1024,0.012631110846996307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,768,0.010537777509954242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,768,0.010252444280518426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,768,0.012267555627557965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,512,0.009581333233250512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,512,0.00811111099190182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,512,0.01096088853147295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,256,0.00832088871134652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,256,0.006180444525347815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,256,0.008586666650242275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,128,0.004648000001907349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,128,0.008162666526105668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,64,0.00479911102188958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,64,0.008716444174448649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,16384,32,0.004920000003443824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,16384,32,0.009626666704813639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,65536,0.24707200792100695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,65536,0.4227920108371311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,16384,128,0.006117333140638139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,65536,0.2973324457804362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,16384,0.10287733210457696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,16384,0.07263377639982435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,16384,0.06331555710898505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,12288,0.04911999901135763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,12288,0.07807466718885633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,12288,0.0528657767507765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,10240,0.06671288940641615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,10240,0.043450666798485645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,10240,0.0425146652592553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,8192,0.03547199898295932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,8192,0.053542223241594106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,8192,0.03680088784959581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,7168,0.03213333421283298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,7168,0.0491848886013031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,7168,0.03251911203066508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,6144,0.02860800094074673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,6144,0.04160177707672119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,6144,0.02882222334543864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,5120,0.025058666865030926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,5120,0.03593600127432082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,5120,0.025419554776615564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,4096,0.02120888895458645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,4096,0.02918755677011278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,4096,0.0214355554845598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,3584,0.019524445136388142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,3584,0.026479111777411565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,3584,0.019753777318530612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,3072,0.017841777867741056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,3072,0.025389333566029865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,3072,0.0177831103404363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,2560,0.016224000189039443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,2560,0.020433778564135235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,2560,0.01608088943693373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,2048,0.016882666283183627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,2048,0.014460444450378418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,1536,0.012491555677519904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,1536,0.0144213338692983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,2048,0.01426488823360867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,1536,0.01257599973016315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,1024,0.009933333430025313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,1024,0.0105608892109659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,1024,0.011233777635627322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,768,0.009542221824328104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,768,0.00849955528974533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,768,0.010654222634103564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,512,0.008573333422342936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,512,0.00703111125363244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,512,0.0087333329849773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,256,0.006256888724035687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,256,0.004810666872395409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,256,0.007696888513035244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,12288,128,0.0053048887186580235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,128,0.004063999901215236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,128,0.007370666497283512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,64,0.0038888889054457345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,64,0.007841777470376756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,12288,32,0.00397511116332478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,12288,32,0.008408000071843466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,65536,0.3608880043029785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,65536,0.22044800387488472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,16384,0.05661866399976942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,65536,0.26750755310058594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,16384,0.09006755881839329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,16384,0.07021600008010864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,12288,0.044156445397271045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,12288,0.06761688656277128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,12288,0.05089155501789517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,10240,0.03830666674507989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,10240,0.05764444669087728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,10240,0.04370577798949348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,8192,0.03222489025857713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,8192,0.04733333322736952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,8192,0.03575111097759671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,7168,0.029049777322345312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,7168,0.04095288780000474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,7168,0.031217777066760596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,6144,0.025747555825445387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,6144,0.035790221558676824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,6144,0.028246220615175035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,5120,0.030960887670516968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,5120,0.024433778391944036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,4096,0.019652444455358718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,4096,0.026259554757012263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,5120,0.022563555174403723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,4096,0.020649777518378366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,3584,0.02316444449954563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,3584,0.019261333677503798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,3072,0.016195555528004963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,3584,0.017768000562985737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,3072,0.020488000578350492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,3072,0.017363554901546903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,2560,0.014849777022997538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,2560,0.017337777548366122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,2560,0.015652444627549913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,2048,0.014392889208263822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,2048,0.014305777019924588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,2048,0.013134222063753339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,1536,0.011491555306646558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,1536,0.011391110718250275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,1536,0.012493333054913415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,1024,0.009762666291660732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,1024,0.008832889298597971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,1024,0.011003555523024665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,768,0.008984888593355814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,768,0.00757155567407608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,768,0.010259555445777046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,512,0.007642666498819987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,512,0.006360000206364526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,512,0.008257778154479133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,256,0.004313777718279096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,256,0.005636444522274866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,256,0.007674666742483775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,10240,128,0.005136000199450387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,128,0.0037537775933742523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,128,0.007258666886223688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,64,0.00342399999499321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,64,0.007608888877762689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,10240,32,0.0037662221325768363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,10240,32,0.00795733349190818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,65536,0.16343200206756592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,65536,0.2863271236419678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,65536,0.1976062191857232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,16384,0.04292177822854784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,16384,0.07051644722620647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,16384,0.050435556305779346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,12288,0.033462223079469465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,12288,0.053652445475260414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,12288,0.03681600093841553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,10240,0.04899911085764567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,10240,0.03206844462288751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,8192,0.02513422237502204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,8192,0.03783288929197524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,10240,0.02942755487230089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,8192,0.027243556247817144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,7168,0.02221244408024682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,7168,0.033055110110176936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,7168,0.024134222004148696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,6144,0.029064890411165025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,6144,0.021765333082940843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,5120,0.017831110292010836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,5120,0.025198222862349615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,5120,0.019519110520680744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,6144,0.020025778147909377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,4096,0.015625778171751235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,4096,0.0210844443904029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,4096,0.016466667254765827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,3584,0.014615999327765571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,3584,0.018811555372344125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,3584,0.015383111106024848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,3072,0.013344888885815939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,3072,0.016674665941132438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,3072,0.014212444424629211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,2560,0.012356444365448423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,2560,0.014476444986131458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,2560,0.013071999781661563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,2048,0.011128889189826118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,2048,0.0120000003112687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,2048,0.011899555722872416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,1536,0.01017066670788659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,1536,0.009918221996890174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,1536,0.01071733319097095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,1024,0.008833777573373582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,1024,0.007879111501905654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,1024,0.009654222263230218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,768,0.008196444147162968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,768,0.006822222222884496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,768,0.008560000194443597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,512,0.006701333241330252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,512,0.0058746664888328975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,512,0.007181333170996771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,256,0.0052533331844541765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,256,0.004047111090686587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,256,0.00675999994079272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,128,0.003710222327046924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,8192,128,0.004907555464241239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,128,0.00654755574133661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,64,0.003579555700222651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,64,0.0070159998204973005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,8192,32,0.003726222034957674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,8192,32,0.007312888900438945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,65536,0.2665546735127767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,65536,0.16026133961147732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,65536,0.18442489041222465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,16384,0.04288977715704176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,16384,0.07015644179450141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,16384,0.04830311073197258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,12288,0.03288888931274414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,12288,0.06007288561926948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,12288,0.036610666248533465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,10240,0.04190399911668566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,10240,0.03081688947147793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,10240,0.028866665230857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,8192,0.034430222378836736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,8192,0.02643022272321913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,7168,0.022110222114457026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,7168,0.030649777915742662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,7168,0.02377333409256405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,6144,0.019648000597953796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,6144,0.026876444617907207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,6144,0.02128444446457757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,5120,0.017775999175177682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,8192,0.024328889118300542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,5120,0.023408000667889912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,5120,0.019143111175960965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,4096,0.015489778584904142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,4096,0.01960711181163788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,4096,0.016378665963808697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,3584,0.014477334088749357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,3584,0.01740800009833442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,3584,0.015198222464985318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,3072,0.01313777764638265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,3072,0.015561777684423657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,3072,0.014081777797804939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,2560,0.012447111308574677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,2560,0.012834666503800286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,2560,0.012984889249006907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,2048,0.0107342220014996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,2048,0.011347555451922946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,2048,0.011848889291286469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,1536,0.010122666756312052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,1536,0.008991111483838823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,1536,0.010592000352011787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,1024,0.009197333620654212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,1024,0.007248000138335758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,1024,0.009547555612193214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,768,0.007779555188284979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,768,0.0063680000603199005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,768,0.007631111476156447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,512,0.005055111315515307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,512,0.006158222340875202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,512,0.007161777880456712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,256,0.005215999980767568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,256,0.003858666867017746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,256,0.006792888873153263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,128,0.003633777714437909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,7168,128,0.004932444542646408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,128,0.006526222245560751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,64,0.0033200000309281876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,64,0.006875555548402999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,7168,32,0.003407111184464561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,7168,32,0.007272889216740926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,65536,0.21811644236246744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,65536,0.18893422020806205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,16384,0.06069778071509468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,16384,0.04672355453173319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,12288,0.03515200151337518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,65536,0.1567146645651923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,16384,0.04483200112978617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,12288,0.048581331968307495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,12288,0.03507377703984579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,10240,0.03026755650838216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,10240,0.03594755464129978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,10240,0.03087200058831109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,8192,0.02540622154871623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,8192,0.03021777669588725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,8192,0.02587911155488756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,7168,0.02274399995803833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,7168,0.02591111097070906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,7168,0.02312711046801673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,6144,0.0203795548942354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,6144,0.022779555784331426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,6144,0.020749333831999037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,5120,0.01811911165714264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,5120,0.019876443677478366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,5120,0.018698667486508686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,4096,0.015670221712854173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,4096,0.017254221770498488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,4096,0.015984000431166757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,3584,0.01462222304609087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,3584,0.015040889382362366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,3584,0.015105777316623263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,3072,0.013400889105266996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,3072,0.012977778083748288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,3072,0.013923555612564087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,2560,0.012281777958075205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,2560,0.011431111229790581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,2560,0.012839111189047495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,2048,0.010948444406191507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,2048,0.009732444253232744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,2048,0.011600889265537262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,1536,0.009734222458468543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,1536,0.008239111138714684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,1536,0.010478222535716163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,1024,0.008250666989220513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,1024,0.006751110984219446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,1024,0.008458666503429413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,768,0.006632888896597757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,768,0.005980444451173146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,768,0.00757422215408749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,512,0.005473777651786804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,512,0.004530666602982415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,512,0.007176888485749562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,256,0.005098666581842634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,256,0.0037679999238914917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,256,0.006792000184456508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,6144,128,0.0047564444442590075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,128,0.0034222222036785553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,128,0.006542222367392645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,64,0.0032675556010670135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,64,0.006879999819729064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,6144,32,0.0034186666210492453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,6144,32,0.007071110937330458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,65536,0.12484444512261285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,65536,0.1882257726457384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,65536,0.1751128832499186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,16384,0.034811556339263916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,16384,0.05391377872890896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,16384,0.045048889186647206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,12288,0.04235733217663235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,12288,0.03376266691419814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,10240,0.024113777610990737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,12288,0.027444443768925134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,10240,0.037104000647862755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,10240,0.0296382208665212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,8192,0.020536889632542927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,8192,0.025817778375413682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,8192,0.02514311174551646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,7168,0.018739556272824604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,7168,0.02340000040001339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,7168,0.02275644408331977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,6144,0.01667644414636824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,6144,0.020654223031467862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,6144,0.021060445242457922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,5120,0.014979556202888489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,5120,0.018068444397714403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,5120,0.01848355597919888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,4096,0.013070222404268054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,4096,0.015236443943447538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,4096,0.01572088897228241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,3584,0.012372444073359171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,3584,0.013493333425786761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,3584,0.014652444256676568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,3072,0.011552000211344825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,3072,0.01200177768866221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,3072,0.013771555489963956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,2560,0.01075466639465756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,2560,0.010102222363154093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,2560,0.012535110943847232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,2048,0.00982844415638182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,2048,0.008842666943868002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,2048,0.011419555379284753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,1536,0.007521777517265744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,1536,0.008912000391218398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,1536,0.010471111370457543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,1024,0.00720266666677263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,1024,0.006183111005359226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,1024,0.008053333395057255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,768,0.006100444330109491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,768,0.005474666754404704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,768,0.007626666790909237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,512,0.005526222288608551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,512,0.0041200001206662925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,512,0.007040888898902469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,256,0.005035555611054103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,256,0.003641777982314428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,256,0.006788444601827198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,5120,128,0.0046568889584806235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,128,0.0033608888172441055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,128,0.006467555546098285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,64,0.003132444495956103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,64,0.006781333436568578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,5120,32,0.0032471112079090546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,5120,32,0.0069120000633928515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,65536,0.098471999168396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,65536,0.14874577522277832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,65536,0.10480444298850165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,16384,0.028196444114049275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,16384,0.042345778809653394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,16384,0.028976887464523315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,12288,0.02206666602028741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,12288,0.033999999364217125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,12288,0.0229120006163915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,10240,0.01943733294804891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,10240,0.030279109875361126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,10240,0.020599999361568026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,8192,0.026112889250119526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,8192,0.017998221847746108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,7168,0.02366133365366194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,8192,0.017095999585257638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,7168,0.01570044457912445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,7168,0.016325333052211337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,6144,0.01422933406300015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,6144,0.01864622202184465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,6144,0.015093333191341825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,5120,0.016379555066426594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,5120,0.013901333014170328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,5120,0.012838222086429596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,4096,0.011643555429246692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,4096,0.014547554983033074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,4096,0.012096888489193387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,3584,0.01105955574247572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,3584,0.01293511109219657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,3584,0.011559999651379056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,3072,0.010254222485754225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,3072,0.011226666470368704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,3072,0.010946667028797997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,2560,0.009722666607962714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,2560,0.010063110954231685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,2560,0.01037688884470198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,2048,0.008712000317043727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,2048,0.00959288908375634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,2048,0.008964444200197855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,1536,0.007540444533030192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,1536,0.008058666355080074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,1536,0.00812533332241906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,1024,0.006198222024573221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,1024,0.005928888916969299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,1024,0.00675733346078131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,768,0.005597333527273602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,768,0.004957333207130432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,768,0.00654666663871871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,512,0.0052435555391841466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,512,0.0041475556790828705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,512,0.00627377794848548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,256,0.004863111095296012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,256,0.0036799998746977914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,256,0.006008000009589725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,128,0.003411555455790626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,128,0.005792888916201062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,4096,128,0.004544888933499654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,64,0.0032044444233179092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,64,0.006144888699054718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,4096,32,0.003297777846455574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,4096,32,0.0062871111763848206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,65536,0.134716444545322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,65536,0.09524089097976685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,65536,0.0979200005531311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,16384,0.02776977750990126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,16384,0.040397332774268255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,16384,0.028001778655582007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,12288,0.02174844510025448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,12288,0.032067557175954185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,12288,0.021968000464969214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,10240,0.019457777341206867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,10240,0.028414222929212783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,10240,0.019684443871180218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,8192,0.016693333784739178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,8192,0.0199955552816391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,8192,0.01703555550840166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,7168,0.015537778536478678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,7168,0.018053332964579265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,7168,0.01573333309756385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,6144,0.013912889692518445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,6144,0.01594311164485084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,6144,0.01462222304609087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,5120,0.012940444052219391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,5120,0.013809777796268463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,5120,0.013515555196338229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,4096,0.011407111254003314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,4096,0.011887999872366587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,4096,0.011928000383906893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,3584,0.010904889139864178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,3584,0.010537777509954242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,3584,0.011316444310877057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,3072,0.010114666488435533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,3072,0.009661333428488838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,3072,0.010680889089902243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,2560,0.009647111097971598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,2560,0.009039999710188972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,2560,0.010115555591053432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,2048,0.007925333248244392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,2048,0.008811555802822113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,2048,0.009328000247478485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,1536,0.007239110767841339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,1536,0.006970666348934174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,1536,0.007640889121426477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,1024,0.005982222242487802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,1024,0.0058124446206622655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,1024,0.006778666542636023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,768,0.005523555394675996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,768,0.004868444469239977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,768,0.00654755574133661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,512,0.005211555709441503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,512,0.0042257776690853965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,512,0.00629066675901413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,256,0.004783111313978831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,256,0.0036159998012913596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,256,0.005995555470387141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3584,128,0.004522666749027041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,128,0.00332622230052948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,128,0.0057608890864584185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,64,0.0031715554909573663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,64,0.006052444378534953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3584,32,0.003248888792263137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3584,32,0.006114666660626729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,65536,0.11815110842386882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,65536,0.09295911259121364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,65536,0.08695377906163533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,16384,0.02737422287464142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,16384,0.03529333406024509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,16384,0.026055110825432673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,12288,0.021371554997232225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,12288,0.027726221415731642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,12288,0.021273778544531927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,10240,0.018903111418088276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,10240,0.024558222956127588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,10240,0.01921422282854716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,8192,0.016427555018001132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,8192,0.021750221649805706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,8192,0.01660711069901784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,7168,0.015006222658687167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,7168,0.01928622192806668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,7168,0.015245333313941956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,6144,0.013612444202105204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,6144,0.014885332849290637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,6144,0.01425244410832723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,5120,0.012502222425407834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,5120,0.012972444295883179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,5120,0.013207111093733044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,4096,0.011179555621412067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,4096,0.011131555669837527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,4096,0.011493333511882357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,3584,0.010562666588359408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,3584,0.010235555469989777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,3584,0.0110880004035102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,3072,0.009849777652157677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,3072,0.0092284447617001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,3072,0.010523555179437002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,2560,0.009277332988050249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,2560,0.008424888882372115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,2560,0.009917333722114563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,2048,0.007975111405054728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,2048,0.007477333148320516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,2048,0.008800889054934183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,1536,0.006667555620272954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,1536,0.006681777950790193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,1536,0.007167110840479533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,1024,0.005704888867007361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,1024,0.005169777820507686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,1024,0.006657777975002925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,768,0.005437333136796951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,768,0.004375111311674118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,768,0.00647288892004225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,512,0.0052133335007561584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,512,0.0038808890514903595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,512,0.006253333141406377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,256,0.004741333425045013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,256,0.0034462221794658234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,256,0.005947555518812603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,3072,128,0.004472889006137848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,128,0.003198222153716617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,128,0.005764444255166584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,64,0.0030097777230872046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,64,0.005936000082227919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,3072,32,0.003129777808984121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,3072,32,0.0060159998635451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,65536,0.09036533037821452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,65536,0.09884177976184422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,65536,0.07956177658504911
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,16384,0.02719199988577101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,16384,0.030922667847739324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,16384,0.02498666610982683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,12288,0.02106133268939124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,12288,0.02522755497031742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,12288,0.020610666937298246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,10240,0.018633777896563213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,10240,0.022472888231277466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,10240,0.018434666925006442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,8192,0.015955555770132277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,8192,0.019357333580652874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,8192,0.015899555550681222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,7168,0.014959999256663851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,7168,0.017528888252046373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,7168,0.014828445182906257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,6144,0.013423999978436364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,6144,0.015929778416951496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,6144,0.014004444082578024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,5120,0.012255111502276527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,5120,0.014324444863531323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,5120,0.012845333251688214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,4096,0.010935111178292168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,4096,0.010685332947307162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,4096,0.011321778098742167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,3584,0.00979911122057173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,3584,0.010453333457310995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,3584,0.010785777535703448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,3072,0.00959822204377916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,3072,0.009100444614887238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,3072,0.010267555713653564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,2560,0.008608889248636033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,2560,0.008153777983453538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,2560,0.009564444422721863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,2048,0.0075093333919843035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,2048,0.007330666813585494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,2048,0.007892444729804993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,1536,0.006456888798210356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,1536,0.007068444457319047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,1024,0.00565422202150027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,1024,0.005032888717121548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,1536,0.0064995553758409284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,1024,0.006629333313968446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,768,0.004503111044565837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,768,0.0063786668082078295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,512,0.005079999979999331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,512,0.003924444317817688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,512,0.006203555398517185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,256,0.003400000019205941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,256,0.0058764442801475525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,768,0.005398222141795688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,128,0.005655999812814925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,256,0.0046942221621672315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2560,128,0.004447999927732679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,128,0.003137777869900068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,64,0.0029537777105967202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,64,0.0058462222417195635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2560,32,0.003039111072818438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2560,32,0.005927111125654644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,65536,0.08072711361779107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,65536,0.08044444190131293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,65536,0.07466222180260552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,16384,0.02458222210407257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,16384,0.025812443759706285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,16384,0.02402666707833608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,12288,0.01924088928434584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,12288,0.02092711130777995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,12288,0.0196079992585712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,10240,0.01862044466866387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,10240,0.016903999778959487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,10240,0.017782222893502977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,8192,0.0147724449634552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,8192,0.01642933322323693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,8192,0.015306666493415833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,7168,0.013807999591032663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,7168,0.01571377780702379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,7168,0.014479110638300577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,6144,0.012455110748608908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,6144,0.014307555225160388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,6144,0.013431111143694984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,5120,0.011495999991893768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,5120,0.013245333400037555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,5120,0.012505778008037143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,4096,0.010302222437328762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,4096,0.012088889049159156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,4096,0.01108088923825158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,3584,0.009766221874290043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,3584,0.009696889254781935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,3584,0.0105262224872907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,3072,0.00850133349498113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,3072,0.009967111051082611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,2560,0.007723555796676212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,2560,0.007892444729804993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,2560,0.00849866701496972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,2048,0.007187555233637492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,2048,0.007419555551475949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,1536,0.006179555422729916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,1536,0.005896889087226655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,1536,0.00701155596309238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,3072,0.00889866633547677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,2048,0.006940444310506185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,1024,0.005543111099137201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,1024,0.0046160001721647055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,1024,0.006509333435032103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,768,0.0053102220926019884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,768,0.0041884444653987885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,768,0.0062977779242727495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,512,0.005068444543414646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,512,0.0037431112594074676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,512,0.006110222389300664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,256,0.003321777822242843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,256,0.005853333406978183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,128,0.0030373332815037835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,128,0.00563733321097162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,64,0.002891555635465516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,64,0.005771555420425203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,2048,32,0.002959111084540685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,2048,32,0.005771555420425203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,256,0.004634666774008009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,65536,0.079930669731564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,2048,128,0.004387555436955558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,65536,0.061218665705786816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,65536,0.07064355744255914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,16384,0.02408711115519206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,16384,0.02079288827048408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,16384,0.02313599983851115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,12288,0.01904266741540697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,12288,0.016805332567956712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,12288,0.019283556275897555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,10240,0.016743110285864938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,10240,0.014927110738224454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,10240,0.017295999659432303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,8192,0.014137778017255994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,8192,0.0151573336786694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,7168,0.013449778159459433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,7168,0.012873777912722694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,7168,0.014127110441525778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,6144,0.012289777398109436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,8192,0.014518222875065274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,6144,0.012191111014948951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,6144,0.013230221966902414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,5120,0.011242667006121742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,5120,0.011206222077210745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,5120,0.012353777885437012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,4096,0.010303110712104374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,4096,0.01075111081202825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,3584,0.009098666409651438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,3584,0.00903555585278405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,3584,0.009707555174827576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,4096,0.009841777384281158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,3072,0.008114666574531132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,3072,0.008357333640257517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,3072,0.008750222623348236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,2560,0.007221333682537079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,2560,0.0075502221783002215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,2560,0.007967111137178209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,2048,0.006506666541099548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,2048,0.00739555557568868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,1536,0.006088888893524806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,1536,0.005240889059172736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,1536,0.006976889239417181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,1024,0.004383999854326248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,2048,0.006585777633719974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,1024,0.006481777876615524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,1024,0.005472888963090048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,768,0.005199999858935674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,768,0.004071111066473855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,768,0.0063235556913746735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,512,0.004943111290534337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,256,0.0033235556135574975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,512,0.003699555579158995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,512,0.006083555519580841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,256,0.004605333424276776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,256,0.005790222022268508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1536,128,0.004279110994603899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,128,0.003010666618744532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,128,0.005611555443869696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,64,0.0028577778074476454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,64,0.005680000202523337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1536,32,0.002971555623743269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1536,32,0.0057555557125144536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,65536,0.07614399989446004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,65536,0.04318933354483711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,16384,0.022916444473796423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,65536,0.06955111026763916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,16384,0.01702666613790724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,12288,0.014142221874660917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,16384,0.02271466619438595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,12288,0.01868533260292477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,10240,0.012693333129088083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,10240,0.016938666502634685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,8192,0.013783111340469785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,8192,0.011353777514563667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,8192,0.0147706667582194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,7168,0.012776888906955719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,7168,0.010832000109884473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,10240,0.015827554795477126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,12288,0.018022222651375663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,7168,0.013910222384664746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,6144,0.011658666862381829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,6144,0.01015911085738076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,6144,0.012784000072214337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,5120,0.009696000152164036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,5120,0.009993777506881291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,5120,0.011608888705571493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,4096,0.008810666700204214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,4096,0.008942222429646386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,4096,0.00945955514907837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,3584,0.008201777935028076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,3584,0.008791111409664154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,3584,0.008792888787057664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,3072,0.007466666400432587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,3072,0.008452444440788692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,3072,0.008475555313958062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,2560,0.006904888898134232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,2560,0.006953777538405524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,2560,0.007898666792445714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,2048,0.006395555618736479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,2048,0.0058871110280354815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,2048,0.007326222128338284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,1536,0.005949333310127258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,1536,0.005108444641033809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,1536,0.006856888946559694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,1024,0.005373333477311664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,1024,0.004314666820896996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,1024,0.0064151109092765385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,768,0.005062222066852781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,768,0.003965333518054751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,768,0.006206222292449739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,512,0.004917333523432414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,512,0.0036355555057525635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,512,0.005949333310127258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,256,0.004512888689835866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,256,0.0032524443748924467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,256,0.005756444401211209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,1024,128,0.00433688900536961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,128,0.002977777686383989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,128,0.005622222191757626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,64,0.0028293333533737394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,64,0.005649777750174205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,1024,32,0.0028826666788922418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,1024,32,0.00566133318675889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,65536,0.07585599687364367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,65536,0.037540445725123085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,65536,0.06982666916317411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,16384,0.0224906669722663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,16384,0.014263111684057446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,16384,0.022686221533351477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,12288,0.01779822177357144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,12288,0.012042666474978128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,12288,0.018720888429217868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,10240,0.015728889240158927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,10240,0.010966222319338055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,10240,0.016930666234758165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,8192,0.013636444177892474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,8192,0.010002666877375709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,8192,0.014663111832406787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,7168,0.012487110992272695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,7168,0.009575111170609793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,7168,0.0137155552705129
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,6144,0.010808889236715106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,6144,0.00906311141120063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,6144,0.012519111235936483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,5120,0.009731555150614845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,5120,0.009923555784755284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,5120,0.010788444843557147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,4096,0.008412444757090675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,4096,0.009029332962301042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,4096,0.009114666945404476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,3584,0.007923555870850882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,3584,0.00832088871134652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,3584,0.008824889030721452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,3072,0.00741155528359943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,3072,0.00775822252035141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,3072,0.008346666892369589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,2560,0.00685511115524504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,2560,0.0063733334342638654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,2560,0.007869333028793335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,2048,0.006329777754015393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,2048,0.00566133318675889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,2048,0.007273777491516537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,1536,0.005867555737495422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,1536,0.004958222309748332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,1536,0.00685422205262714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,1024,0.005256888767083486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,1024,0.004252444538805219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,1024,0.00638488887084855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,768,0.005014222115278244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,768,0.0038746665749284956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,768,0.006178666734033161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,512,0.004794666750563515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,512,0.0035208890007601846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,512,0.006017777654859755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,256,0.0031386667655573953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,256,0.005718222094906702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,128,0.00423911131090588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,128,0.0028977776981062363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,128,0.005555555638339784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,768,256,0.004476444588767158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,64,0.0027804445061418745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,64,0.005648889061477449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,768,32,0.002809777855873108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,768,32,0.005601777798599667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,65536,0.031352887550989784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,65536,0.06989955239825778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,16384,0.022304889228608873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,16384,0.012010667059156628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,16384,0.02258577777279748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,12288,0.017825777331988018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,12288,0.010568000376224518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,12288,0.018616000811258953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,10240,0.015641777051819693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,10240,0.010052444206343757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,65536,0.07601866457197401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,10240,0.016744888491100736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,8192,0.012999110751681857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,8192,0.010173333187898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,8192,0.014336888988812765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,7168,0.01183199965291553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,7168,0.009798222117953831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,7168,0.012609778179062737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,6144,0.010473777850468954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,6144,0.009219555391205682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,6144,0.0116462219092581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,5120,0.008039111064540016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,5120,0.010453333457310995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,4096,0.008295111358165741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,5120,0.00942133367061615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,4096,0.007249777515729268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,4096,0.009045333498054082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,3584,0.007797333101431529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,3584,0.0074933336840735535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,3584,0.008667555948098501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,3072,0.007328888608349695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,3072,0.007088888850477006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,3072,0.008153777983453538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,2560,0.006810666786299811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,2560,0.0063546668324205614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,2560,0.007740444607204861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,2048,0.0058293334311909145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,2048,0.00618488879667388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,2048,0.007239110767841339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,1536,0.005793777604897817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,1536,0.004934222333961063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,1536,0.006789333290523953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,1024,0.00518222235971027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,1024,0.0042568888101312844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,1024,0.006285333385070165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,768,0.004909333255555895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,768,0.003988444391224119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,768,0.0061048890153567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,512,0.0036026665733920205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,512,0.005902222047249476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,256,0.004361777669853634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,512,0.0046951112647851305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,256,0.003273777870668305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,256,0.005647111270162795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,512,128,0.0041751112374994485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,128,0.0029315555261241067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,128,0.005548444473081165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,64,0.002765333279967308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,64,0.005603555589914322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,512,32,0.002871111035346985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,512,32,0.005550222264395819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,65536,0.02132622235351139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,65536,0.07565777831607394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,65536,0.06946400139066908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,16384,0.010422222316265106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,16384,0.022297778063350256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,12288,0.017312889297803242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,12288,0.009468444519572789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,16384,0.022181333767043218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,12288,0.018142221702469718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,10240,0.014649778604507446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,10240,0.008831111093362173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,10240,0.015963556038008798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,8192,0.01239111108912362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,8192,0.007679999702506595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,8192,0.013030221892727746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,7168,0.011414222419261932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,7168,0.007344889144102733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,7168,0.01219733307758967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,6144,0.010355555348926121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,6144,0.006816888848940532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,6144,0.010997333460383944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,5120,0.00943199959066179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,5120,0.00720088928937912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,5120,0.010206222534179688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,4096,0.008249777886602614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,4096,0.006775110960006714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,4096,0.009097778134875828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,3584,0.007793777518802219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,3584,0.007143999967310164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,3584,0.008712000317043727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,3072,0.007286222444640265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,3072,0.006863111009200414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,3072,0.00814133303032981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,2560,0.006754666566848755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,2560,0.006171555568774541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,2560,0.007687110867765214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,2048,0.006175999840100606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,2048,0.005483555710977978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,2048,0.007191999918884701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,1536,0.005708444449636672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,1536,0.004825777891609403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,1536,0.00675999994079272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,1024,0.005091555416584015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,1024,0.004096888833575779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,1024,0.00628977765639623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,768,0.0037564444873068067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,768,0.006081777728266186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,512,0.0046302220887608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,512,0.0033893332713180115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,512,0.005789333333571752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,256,0.004324444466167026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,256,0.0030257778449190986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,768,0.004861333303981357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,256,0.005607999861240387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,256,128,0.0041511112617121804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,128,0.0028657778683635923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,128,0.005472888963090048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,64,0.0026604444202449587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,64,0.00552977787123786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,256,32,0.0027280000762807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,256,32,0.005482666608360078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,65536,0.07562933365503947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,65536,0.01680444512102339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,65536,0.06939644283718534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,16384,0.00850133349498113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,16384,0.022127999199761286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,12288,0.007359999749395583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,12288,0.016698666744761996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,10240,0.014474666780895658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,10240,0.007124444676770105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,10240,0.014540443817774454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,16384,0.021418665846188862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,12288,0.01682933337158627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,8192,0.01240888900227017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,8192,0.006991110742092133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,8192,0.012390221986505719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,7168,0.011384000380833944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,7168,0.0069306666652361555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,7168,0.011692444483439127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,6144,0.006679111056857639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,6144,0.010285333626800114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,6144,0.010848888920413123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,5120,0.009333333207501305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,5120,0.007060444189442529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,5120,0.010051555931568146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,4096,0.008244444098737504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,4096,0.006617777877383762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,4096,0.00887111077706019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,3584,0.00777511133088006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,3584,0.007133333219422235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,3584,0.008558221989207799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,3072,0.007232888705200619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,3072,0.006761777732107375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,3072,0.008055110772450766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,2560,0.006700444552633498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,2560,0.00609688874748018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,2560,0.007639110916190677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,2048,0.006119110931952794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,2048,0.00544177782204416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,2048,0.007096000015735626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,1536,0.005704000178310607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,1536,0.00479911102188958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,1536,0.006700444552633498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,1024,0.005056888692908817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,1024,0.004090666770935059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,1024,0.006264888991912206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,768,0.004813333352406819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,768,0.003663111064169142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,768,0.006051555689838197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,512,0.004531555705600315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,512,0.0033857776886887024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,512,0.0058204444746176405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,256,0.004263111286693149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,256,0.003042666655447748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,256,0.005601777798599667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,128,128,128,0.004064889003833135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,128,0.0028400001012616684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,128,0.005472888963090048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,64,0.0026746667507621977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,64,0.005452444569932089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,128,32,0.002672000063790215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,128,32,0.005482666608360078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,65536,0.014182221558358936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,65536,0.06961599985758464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,16384,0.007199111084143321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,16384,0.021789333886570398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,12288,0.006713777780532837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,12288,0.016215110818545025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,10240,0.0069306666652361555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,10240,0.014510222607188754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,8192,0.006776000062624614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,8192,0.012376888758606382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,7168,0.006656888872385025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,7168,0.011687999798191918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,6144,0.00654755574133661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,6144,0.01072355525361167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,5120,0.006966222491529252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,5120,0.01000533335738712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,4096,0.006491555521885554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,4096,0.008774222599135505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,3584,0.0069368887278768755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,3584,0.00851822230550978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,3072,0.006623111251327727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,3072,0.008024888734022776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,2560,0.006088888893524806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,2560,0.007631999750932057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,2048,0.005428444594144821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,2048,0.007034666836261749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,1536,0.0047173334492577445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,1536,0.006671999891599019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,1024,0.004023111114899318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,1024,0.006239999913507038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,768,0.0036524443162812125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,768,0.006041777630647023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,512,0.0033235556135574975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,512,0.005776888794369168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,256,0.002965333354141977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,256,0.005601777798599667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,128,0.002806222273243798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,128,0.005442666510740916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,64,0.0026559999419583213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,64,0.005510222166776657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,64,32,0.0026853332916895547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,64,32,0.005457777943876054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,65536,0.013408000270525614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,65536,0.06965244478649564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,16384,0.006753777878151999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,16384,0.02091466718249851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,12288,0.006539555473460092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,12288,0.016198222835858662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,10240,0.006724444528420766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,10240,0.014241778188281588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,8192,0.006721777634488211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,8192,0.012368888490729861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,7168,0.006548444430033366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,7168,0.011580444044537015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,6144,0.0064275554484791225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,6144,0.010677333507272931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,5120,0.006824000014199152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,5120,0.009872888525327047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,4096,0.006411555740568373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,4096,0.008730666504965888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,3584,0.006843555718660355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,3584,0.008440000315507254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,3072,0.006565333240562015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,3072,0.008086221913496653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,2560,0.006054222169849608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,2560,0.0075297777851422625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,2048,0.005385777602593104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,2048,0.007074666519959767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,1536,0.004690666579537922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,1536,0.006720888945791457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,1024,0.004010666575696734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,1024,0.006233777850866318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,768,0.003636444608370463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,768,0.005982222242487802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,512,0.003310222178697586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,512,0.005751111027267244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,256,0.0029786665820413162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,256,0.00554666668176651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,128,0.0027502222607533136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,128,0.005439110928111606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,64,0.0026088888860411118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,64,0.005384888913896348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,128,32,32,0.002598222138153182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,128,32,32,0.005437333136796951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,16384,0.28138576613532174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,16384,0.5306764708624946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,12288,0.3978995482126872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,12288,0.22071377436319986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,16384,0.5166871282789443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,10240,0.3344275686475966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,10240,0.17551733387841117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,12288,0.38392000728183323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,10240,0.29976267284817165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,8192,0.15198400285508898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,8192,0.26862843831380206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,7168,0.1294364399380154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,8192,0.26062220997280544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,7168,0.23460533883836535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,7168,0.23433865441216362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,6144,0.11714933978186713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,6144,0.20328088601430258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,6144,0.18700177139706084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,5120,0.1704577737384372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,5120,0.17070310645633271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,5120,0.09477066993713379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,4096,0.0813795526822408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,4096,0.13817777898576525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,4096,0.1362773312462701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,3584,0.07444266478220622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,3584,0.12099733617570664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,3584,0.12362666924794515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,3072,0.10613155364990234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,3072,0.06139289008246528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,3072,0.10382666852739121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,2560,0.08883021937476264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,2560,0.084934221373664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,2560,0.05325599842601352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,2048,0.07331911060545179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,2048,0.04445777667893303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,2048,0.06965244478649564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,1536,0.03615199857287937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,1536,0.061793777677747935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,1536,0.052460445298088916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,1024,0.04090133309364319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,1024,0.037920888927247785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,768,0.033506668276256986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,768,0.03087466624048021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,1024,0.027264000640975103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,768,0.023375999596383836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,512,0.019167110323905945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,512,0.02372444503837162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,512,0.025051555699772302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,256,0.014702222413486905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,256,0.015585778488053216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,256,0.021684444612926904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,128,0.009727111293209923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,128,0.020643555455737643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,64,0.007595555649863348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,64,0.021791110436121624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,65536,32,0.007032000356250339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,65536,32,0.024254222710927326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,65536,0.305531554751926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,65536,128,0.012237333589129977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,65536,0.5875093142191569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,65536,0.5014124446445042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,16384,0.07839022080103557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,16384,0.1296773354212443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,16384,0.12448978424072266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,12288,0.05881510840521919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,12288,0.09898489051394993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,12288,0.09055999914805095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,10240,0.08342844247817993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,10240,0.07577600081761678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,8192,0.043218665652804904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,8192,0.0758791102303399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,10240,0.05058755477269491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,8192,0.06507733133104113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,7168,0.03800088829464383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,7168,0.05968888600667318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,7168,0.05635822481579251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,6144,0.03271911210483975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,6144,0.05166666706403097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,6144,0.05002133382691277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,5120,0.028747555282380845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,5120,0.044159111049440175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,5120,0.04313333498107063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,4096,0.03689333465364244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,4096,0.02421599957678053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,4096,0.03685866792996725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,3584,0.022686221533351477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,3584,0.0325217776828342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,3584,0.033096889654795326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,3072,0.02032888929049174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,3072,0.02872977654139201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,3072,0.028024001253975764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,2560,0.024969778127140466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,2560,0.02369244396686554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,2560,0.018238221605618794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,2048,0.015771556231710646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,2048,0.02086488902568817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,2048,0.020442666278945077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,1536,0.0136808885468377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,1536,0.016959110895792644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,1536,0.01679999960793389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,1024,0.012226666841242047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,1024,0.011252444651391772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,1024,0.014055111342006259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,768,0.010280888941552904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,768,0.009892444643709395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,768,0.013067555924256643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,512,0.009275555610656738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,512,0.007784888976150089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,512,0.011583111352390714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,256,0.007164444360468123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,256,0.005848000033034219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,256,0.008441777692900764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,16384,128,0.005702222386995952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,128,0.004269333349333869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,128,0.00814222213294771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,64,0.003858666867017746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,64,0.008835555778609382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,16384,32,0.004050666673315896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,16384,32,0.009462222456932068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,65536,0.41705777910020614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,65536,0.3741724491119385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,65536,0.2280471060011122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,16384,0.11108177900314331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,16384,0.06169688701629639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,16384,0.09592533111572266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,12288,0.04837422238455879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,12288,0.08660888671875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,12288,0.06755466593636407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,10240,0.04180888997183906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,10240,0.07821510897742377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,10240,0.05684355232450697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,8192,0.03489155570665995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,8192,0.053655998574362866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,8192,0.05092888739373949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,7168,0.03179288903872172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,7168,0.04798488815625509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,7168,0.043972445858849414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,6144,0.02809777855873108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,6144,0.04055822226736281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,6144,0.03849066628350152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,5120,0.03489333391189575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,5120,0.03326488865746392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,4096,0.029044446018007066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,4096,0.028347555134031508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,3584,0.026008889079093933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,5120,0.02459377712673611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,4096,0.020911999874644812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,3584,0.019311999281247456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,3584,0.026205332742797002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,3072,0.01754488878779941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,3072,0.02288888891537984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,3072,0.022268444299697876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,2560,0.016069332758585613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,2560,0.019832000136375427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,2560,0.019687111179033916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,2048,0.01404800017674764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,2048,0.01648266613483429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,2048,0.016943110360039603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,1536,0.012304888831244575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,1536,0.013209777573744455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,1536,0.01460444430510203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,1024,0.010326222413116032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,1024,0.00979377743270662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,1024,0.012309333516491784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,768,0.00943022221326828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,768,0.008382221890820397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,768,0.011399110986126794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,512,0.008412444757090675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,512,0.0069982219073507525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,512,0.00962222201956643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,256,0.0046897778908411665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,256,0.007755555212497711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,128,0.004061333421203825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,256,0.0058373332851462895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,12288,128,0.0053342220683892565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,128,0.007265778051482306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,64,0.003630222131808599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,64,0.007844444778230455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,12288,32,0.003677333394686381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,12288,32,0.008231110870838165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,65536,0.20179288917117646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,65536,0.3699084387885199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,65536,0.353383117251926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,16384,0.0997004442744785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,16384,0.08781955639521281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,12288,0.04359022114011976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,16384,0.05531733234723409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,12288,0.07897866434521146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,12288,0.06232533189985487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,10240,0.04083911246723599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,10240,0.06848977671729194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,10240,0.05373866690529717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,8192,0.031714667876561485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,8192,0.055098666085137256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,8192,0.046718223227394946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,7168,0.028798222541809082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,7168,0.049266666173934937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,7168,0.0415013333161672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,6144,0.043776889642079674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,6144,0.0350835555129581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,6144,0.02551022171974182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,5120,0.022283555732833013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,5120,0.03059555424584283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,5120,0.03003822101487054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,4096,0.01888799998495314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,4096,0.025082666012975905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,4096,0.026392888691690233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,3584,0.017624000708262127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,3584,0.022494221727053326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,3584,0.024534222152498033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,3072,0.020055111911561754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,3072,0.016180444094869826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,3072,0.020632000433074105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,2560,0.014602666099866232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,2560,0.01789422167672051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,2560,0.018571555614471436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,2048,0.012936000194814471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,2048,0.014634667171372307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,2048,0.015820443630218506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,1536,0.011337777806652917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,1536,0.011834666960769229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,1536,0.013584888643688627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,1024,0.009730666875839233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,1024,0.009249777429633671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,1024,0.011724444727102915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,768,0.007902222375075022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,768,0.010780444575680627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,512,0.00721688899728987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,512,0.006746666712893381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,512,0.008616888688670265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,768,0.0088764445649253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,256,0.005622222191757626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,256,0.004514666481150521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,256,0.007575111256705389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,128,0.0038906666967603895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,128,0.007231999602582719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,64,0.0035111109415690103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,64,0.0075128889746136135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,10240,32,0.0036195553839206696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,10240,128,0.005120888766315249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,10240,32,0.007982222570313348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,65536,0.16247199641333684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,65536,0.2759608957502577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,16384,0.04413777920934889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,16384,0.0743537810113695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,16384,0.06493510802586873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,65536,0.257567114300198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,12288,0.03385422295994229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,12288,0.058693336115943066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,12288,0.04760622315936618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,10240,0.030890666776233252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,10240,0.050287998384899564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,10240,0.04021155503061082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,8192,0.025787555509143408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,8192,0.04026844435267978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,8192,0.03475733266936408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,7168,0.02225333286656274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,7168,0.03265155686272515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,7168,0.031743112537595965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,6144,0.02881422307756212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,6144,0.026917333404223125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,5120,0.01796266602145301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,5120,0.024911110599835713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,5120,0.023975999818907842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,4096,0.015625778171751235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,4096,0.021373333202468023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,4096,0.020489777127901714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,3584,0.014486221803559197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,3584,0.018704000446531508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,6144,0.020138666033744812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,3584,0.018901333212852478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,3072,0.01329066687160068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,3072,0.016679111454221938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,3072,0.01681600014368693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,2560,0.012221333053376941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,2560,0.014187556174066333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,2560,0.014949333336618213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,2048,0.011711999773979187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,2048,0.011197333534558615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,2048,0.013413333230548434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,1536,0.00979200005531311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,1536,0.011787555283970304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,1024,0.008656000097592672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,1536,0.010152888794740042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,1024,0.007695111135641734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,1024,0.010112000008424124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,768,0.006806222101052602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,768,0.008589333130253686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,512,0.006311111152172089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,512,0.005584888988071018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,768,0.008019555773999956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,512,0.007408888803588019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,256,0.003952888978852166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,256,0.0068488890926043195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,128,0.004882666799757215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,128,0.003587555554178026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,128,0.006515555497672822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,64,0.0033119999700122406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,8192,256,0.005352889084153705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,64,0.006835555450783835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,8192,32,0.0035751110149754416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,8192,32,0.007229333122571309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,65536,0.26001956727769643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,65536,0.15429333845774332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,65536,0.22729778289794922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,16384,0.04301955633693271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,16384,0.06799911128150092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,16384,0.06407555606630114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,12288,0.032538665665520564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,12288,0.046112000942230225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,12288,0.045851555135515004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,10240,0.028569777806599934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,10240,0.03994844357172648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,10240,0.03890577620930142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,8192,0.024246222443050806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,8192,0.03304266598489549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,8192,0.03323377834426032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,7168,0.022231111923853557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,7168,0.028778665595584448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,7168,0.028232001596026953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,6144,0.019518221418062847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,6144,0.025191111697090998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,6144,0.026055999928050574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,5120,0.022071111533376906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,5120,0.022448889083332483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,5120,0.01758399936887953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,4096,0.015426667200194465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,4096,0.019187556372748483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,4096,0.020637333393096924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,3584,0.014397333065668741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,3584,0.016533333394262526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,3584,0.01919644408755832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,3072,0.01316088851955202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,3072,0.014683556225564746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,3072,0.016895999511082966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,2560,0.012375999655988483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,2560,0.012852444416946836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,2560,0.015232000086042615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,2048,0.011168888873524137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,2048,0.01068800025516086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,2048,0.012643555800120035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,1536,0.01000533335738712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,1536,0.009004444711738164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,1536,0.011723555624485016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,1024,0.007116444408893585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,1024,0.009760888914267222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,1024,0.008985777695973715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,768,0.00739822205570009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,768,0.0063546668324205614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,768,0.00812711069981257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,512,0.00498577786816491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,512,0.006059555543793573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,512,0.00740444411834081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,256,0.0038764443662431505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,256,0.006812444577614467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,128,0.004882666799757215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,7168,256,0.005219555563396878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,128,0.003524444583389494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,128,0.006484444356626934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,64,0.0032595555401510666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,64,0.006853333363930385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,7168,32,0.003375111147761345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,7168,32,0.007248889240953658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,65536,0.20857066578335234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,65536,0.1630906661351522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,65536,0.23498490121629503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,16384,0.04446488949987623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,16384,0.06110755602518717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,16384,0.062252442042032875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,12288,0.04862311151292589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,12288,0.04456622070736355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,10240,0.030255999830034044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,10240,0.03559644354714288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,12288,0.03501066565513611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,10240,0.03768622212939792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,8192,0.025404444999165002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,8192,0.02922044528855218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,8192,0.03439733386039734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,7168,0.022823999325434368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,7168,0.025701334079106648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,7168,0.02699733277161916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,6144,0.020259555843141343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,6144,0.022823111878501043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,6144,0.02351377738846673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,5120,0.019768888751665752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,5120,0.022069333328141108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,5120,0.018045332696702745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,4096,0.015726221932305228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,4096,0.0170462214284473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,4096,0.01795111099878947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,3584,0.014567111929257711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,3584,0.014836443795098199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,3584,0.016680000556839835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,3072,0.013352000051074557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,3072,0.013204444613721637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,3072,0.015597333510716757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,2560,0.012410666379663678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,2560,0.01147377739350001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,2560,0.01461955573823717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,2048,0.010878221856223213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,2048,0.009847111172146266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,2048,0.012118221984969245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,1536,0.009762666291660732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,1536,0.008357333640257517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,1536,0.010951999988820819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,1024,0.0074933336840735535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,1024,0.00675733346078131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,1024,0.008801778157552084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,768,0.0063279999627007385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,768,0.00609688874748018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,768,0.00812888890504837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,512,0.005515555540720622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,512,0.004393777913517422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,512,0.007354666789372762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,256,0.005103999955786599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,256,0.003850666599141227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,256,0.006787555499209299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,6144,128,0.0047306666771570844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,128,0.0034844444857703317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,128,0.0064791109826829694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,64,0.003212444484233856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,64,0.00674311113026407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,6144,32,0.0032559999575217566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,6144,32,0.007054222126801808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,65536,0.1456257767147488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,65536,0.18335555659400093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,65536,0.22086311711205375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,16384,0.03660800059636434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,16384,0.05241866575347053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,16384,0.05580621957778931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,12288,0.02914488977856106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,12288,0.04224444429079691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,12288,0.0413991113503774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,10240,0.02422044508987003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,10240,0.03684533304638333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,10240,0.03459644317626953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,8192,0.021495110458797876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,8192,0.031255112753974065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,8192,0.03117777903874715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,7168,0.019187556372748483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,7168,0.028627554575602215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,7168,0.027967111931906805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,6144,0.016673778494199116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,6144,0.025606221622890894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,6144,0.023319111929999456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,5120,0.014797333214018079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,5120,0.01780177818404304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,5120,0.021892444954978094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,4096,0.014929778046078153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,4096,0.013266666895813413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,4096,0.01815022197034624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,3584,0.012288889123333825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,3584,0.013320888910028668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,3584,0.016573333077960543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,3072,0.011457777685589261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,3072,0.01184622198343277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,3072,0.01461777753300137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,2560,0.01072266697883606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,2560,0.010393777655230628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,2560,0.013374221821626028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,2048,0.009871111147933537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,2048,0.009108444054921469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,2048,0.011848889291286469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,1536,0.0088746663596895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,1536,0.007684444387753804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,1536,0.01072266697883606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,1024,0.007000889215204451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,1024,0.006309333360857434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,1024,0.008375999828179678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,768,0.006073777874310811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,768,0.005548444473081165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,768,0.007805333369308048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,512,0.005636444522274866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,512,0.004173333446184794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,512,0.007256888680987888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,256,0.005007110950019625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,256,0.0037120001183615792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,256,0.00674044465025266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,5120,128,0.004666666603750653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,128,0.0033928888539473214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,128,0.00638755535085996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,64,0.0031235555393828284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,64,0.006732444382376141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,5120,32,0.003253333270549774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,5120,32,0.006895111252864202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,65536,0.14141066869099936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,65536,0.12470400333404541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,65536,0.12374311023288304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,16384,0.0425946679380205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,16384,0.037639111280441284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,16384,0.03424177898301019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,12288,0.03447644578086005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,12288,0.025791111919615004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,12288,0.027572444743580286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,10240,0.021839110387696162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,10240,0.029539555311203003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,10240,0.024101333485709295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,8192,0.019807111885812547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,8192,0.02539199921819899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,8192,0.02092355489730835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,7168,0.01749955614407857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,7168,0.023201778531074524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,7168,0.019169777631759644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,6144,0.015615999698638916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,6144,0.021049777666727703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,6144,0.01719466679626041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,5120,0.018819555640220642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,5120,0.015456888410780164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,4096,0.01237155579858356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,5120,0.01386044422785441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,4096,0.015631111131774057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,4096,0.013491555220550962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,3584,0.011364444262451597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,3584,0.013631111217869652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,3584,0.012719111310111152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,3072,0.010374222364690568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,3072,0.011937778029176923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,3072,0.011899555722872416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,2560,0.0096124443742964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,2560,0.009072889056470659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,2560,0.010999999940395355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,2048,0.008933333059151968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,2048,0.007886221839321984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,2048,0.009941333697901832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,1536,0.0074417781498697065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,1536,0.0069164443347189165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,1536,0.008440888590282863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,1024,0.00600000015563435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,1024,0.005773333211739858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,1024,0.007266666326257918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,768,0.005559111220969095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,768,0.0046924443708525766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,768,0.0068195557428730865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,512,0.005344888816277186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,512,0.003956444561481476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,512,0.0064026667839950984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,256,0.0035955554081334006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,256,0.006027555714050929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,128,0.00452622233165635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,4096,256,0.00490133340160052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,128,0.003297777846455574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,128,0.0057911111248864075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,64,0.003064888839920362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,64,0.005997333261701796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,4096,32,0.0033324443631701996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,4096,32,0.006199999815887875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,65536,0.12700355052947998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,65536,0.12105688783857559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,16384,0.037535998556349016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,65536,0.11915289031134711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,16384,0.035565333233939275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,12288,0.02496888902452257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,12288,0.029988444513744775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,12288,0.026918222506841023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,16384,0.03162844313515557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,10240,0.02658133374320136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,10240,0.023636443747414485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,10240,0.02052088909678989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,8192,0.01867911054028405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,8192,0.023195554812749226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,8192,0.01995733380317688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,7168,0.016872000363137987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,7168,0.020890666378868952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,7168,0.01816177699300978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,6144,0.014650666051440768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,6144,0.01887911061445872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,6144,0.01628177695804172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,5120,0.016903999778959487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,5120,0.014848889576064216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,4096,0.014979556202888489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,4096,0.013006221916940479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,3584,0.010876444478829702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,5120,0.012902222573757172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,4096,0.01181333346499337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,3584,0.013895110951529609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,3584,0.012198222180207571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,3072,0.010470222267839642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,3072,0.01129155523247189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,2560,0.009538667069541084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,2560,0.00851822230550978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,3072,0.010102222363154093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,2560,0.010630221830474006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,2048,0.008611555728647444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,2048,0.0075191110372543335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,2048,0.009383999639087254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,1536,0.0070462218589252895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,1536,0.006743999818960826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,1536,0.007913777397738563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,1024,0.00572622236278322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,1024,0.007048000064161088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,1024,0.005976000179847081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,768,0.005571555760171678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,768,0.004383111165629493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,768,0.00665155549844106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,512,0.0053022222386466126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,512,0.0038720000949170855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,512,0.006298666612969504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,256,0.004858666823969947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,256,0.0034231110993358824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,256,0.005951111101441913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3584,128,0.004564444637960858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,128,0.0031555555760860443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,128,0.0057546666098965546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,64,0.0030168888883458245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,64,0.005941333456171884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3584,32,0.0030737777964936364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3584,32,0.006088888893524806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,65536,0.11889421939849854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,16384,0.03049066662788391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,16384,0.0358862214618259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,65536,0.11217955748240153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,16384,0.032941334777408175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,12288,0.024498666326204937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,12288,0.027186666925748188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,12288,0.024689777029885188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,10240,0.023997333314683702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,65536,0.11866311232248943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,10240,0.0218213333023919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,10240,0.02055022286044227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,8192,0.017927999297777813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,8192,0.021249777740902368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,8192,0.018776888648668926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,7168,0.016553777787420485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,7168,0.018945778409639995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,7168,0.017177777157889474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,6144,0.017101332545280457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,6144,0.01570755574438307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,6144,0.014464888307783337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,5120,0.015423999892340766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,5120,0.014302222265137566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,4096,0.013835555149449242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,4096,0.01236355553070704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,3584,0.009886222581068674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,5120,0.012830221818553077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,3584,0.01163111130396525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,3072,0.009877333210574256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,3072,0.00907911111911138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,3072,0.010994666980372535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,2048,0.007340444458855524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,2560,0.00924533357222875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,2560,0.008249777886602614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,2560,0.010281778044170803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,4096,0.011333333121405708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,2048,0.007821333077218797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,2048,0.00908799966176351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,1536,0.006617777877383762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,1536,0.006599999964237213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,3584,0.010750222537252637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,1536,0.0076142218377855085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,1024,0.005798222290145026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,1024,0.005469333380460739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,1024,0.0069475554757648045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,768,0.005471999860472149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,768,0.004491555607981152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,768,0.006809777683681912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,512,0.005218666460778978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,512,0.0037866665257347953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,512,0.006236444330877728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,256,0.004763555609517627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,256,0.00342399999499321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,256,0.005902222047249476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,3072,128,0.004559111264016894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,128,0.003140444349911478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,128,0.005698666804366642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,64,0.0030008889734745026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,64,0.005886222339338726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,3072,32,0.0030453333424197305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,3072,32,0.006039111150635614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,65536,0.09508799844317967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,65536,0.10812444157070583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,65536,0.09896000226338704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,16384,0.02884711159600152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,16384,0.030589332183202107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,12288,0.02241955531968011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,12288,0.023464888334274292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,12288,0.02273066673013899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,10240,0.018927999668651156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,10240,0.020614221692085266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,10240,0.019925332731670804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,8192,0.016905777984195285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,8192,0.017846221725145977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,8192,0.017495999733606975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,7168,0.015416888727082146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,7168,0.01640888883007897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,7168,0.015594666202863058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,16384,0.029658665259679157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,6144,0.014835556348164877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,6144,0.013872000078360239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,6144,0.014428445034556918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,5120,0.012151111331250934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,5120,0.01349688900841607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,5120,0.013273778061072031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,4096,0.01183999992079205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,4096,0.011571555501884885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,3584,0.010327999790509542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,3584,0.010567111273606619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,3584,0.010824888944625854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,3072,0.009576000273227692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,3072,0.009686222506894005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,3072,0.010272888673676385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,2560,0.008603555460770925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,2560,0.007975999679830339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,2560,0.009391999906963771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,2048,0.007423111134105259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,2048,0.007374222079912822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,2048,0.008006222546100616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,1536,0.00646577775478363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,1536,0.0061919999619325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,1536,0.007309333317809635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,1024,0.005006222261322869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,1024,0.00674044465025266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,4096,0.011049778097205691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,1024,0.0057813334796163775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,768,0.005422222117582957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,768,0.004256000121434529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,768,0.006451555424266391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,512,0.005085333353943295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,512,0.0038568890757030914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,512,0.006198222024573221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,256,0.004754666652944353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,256,0.003391999958289994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,256,0.0058604445722368025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2560,128,0.004504888835880491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,128,0.003126222226354811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,128,0.005615111026499007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,64,0.0029680000411139596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,64,0.0058106668293476105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2560,32,0.003006222140457895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2560,32,0.005852444304360284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,65536,0.10341511170069377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,65536,0.07605599694781832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,16384,0.02479822271400028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,65536,0.0782346659236484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,16384,0.02451466686195797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,12288,0.022408889399634466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,12288,0.020285333196322124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,16384,0.02794755498568217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,12288,0.01980355547534095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,10240,0.018681777848137748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,10240,0.01793688866827223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,10240,0.01789155602455139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,8192,0.016698666744761996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,8192,0.015309333801269531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,8192,0.015490666031837463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,7168,0.015001777145597668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,7168,0.014273777604103088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,7168,0.014448000325096978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,6144,0.013314666847387949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,6144,0.012912888493802814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,6144,0.013421333498424955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,5120,0.011566222541862063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,5120,0.012558221817016602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,4096,0.01055111073785358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,4096,0.011033777561452655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,3584,0.010040000081062317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,3584,0.009191999832789103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,3584,0.010441777606805166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,3072,0.00869511150651508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,3072,0.00850577818022834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,5120,0.011879111329714457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,4096,0.010822222464614444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,3072,0.009266667068004608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,2560,0.007576000359323289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,2560,0.00775733341773351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,2560,0.00831822223133511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,2048,0.006772444479995304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,2048,0.006911110960774952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,2048,0.0076453329788313965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,1536,0.006046222315894232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,1536,0.006171555568774541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,1536,0.007103111180994246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,1024,0.00562755556570159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,1024,0.004492444296677907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,1024,0.006635555376609166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,768,0.005281777845488654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,768,0.004159111115667555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,768,0.0063822223908371395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,512,0.0037973332736227247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,512,0.00507733349998792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,512,0.006126222097211414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,256,0.0046968890560997855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,256,0.0033520000676314035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,256,0.0057591112951437635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,128,0.0031102223114834893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,128,0.005547555370463266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,2048,128,0.004393777913517422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,64,0.0028924445311228433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,65536,0.06216444571812948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,64,0.005694222119119432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,2048,32,0.002951111023624738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,2048,32,0.005731555736727185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,65536,0.08628622028562759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,65536,0.07563466495937772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,16384,0.023777777949968975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,16384,0.020615110794703167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,16384,0.023864888482623633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,12288,0.01867022282547421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,12288,0.016752888758977253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,12288,0.019471999671724107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,10240,0.01676266723208957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,10240,0.014661333627170987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,10240,0.017109332813156974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,8192,0.014532445205582513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,8192,0.012980444563759698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,8192,0.015056888262430826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,7168,0.013305777476893531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,7168,0.012073777616024017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,7168,0.014133334159851074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,6144,0.01221244451072481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,6144,0.011151110960377587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,6144,0.01333333303531011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,5120,0.011066666907734342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,5120,0.01200177768866221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,5120,0.012228444218635559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,4096,0.009447111023796929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,4096,0.00980444418059455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,4096,0.010494222243626913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,3584,0.008830221990744272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,3584,0.009071110851234859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,3584,0.009639110830095079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,3072,0.007807110746701558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,3072,0.008371555142932469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,3072,0.008814222282833522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,2560,0.007119999991522894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,2560,0.0074346669846110875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,2560,0.008073777788215214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,2048,0.006665777828958299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,2048,0.0062977779242727495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,2048,0.0075902218619982404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,1536,0.00612088872326745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,1536,0.005038222091065513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,1536,0.00703999979628457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,1024,0.005566222386227713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,1024,0.00434755575325754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,1024,0.006554666492674086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,768,0.005240889059172736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,768,0.003974222060706881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,768,0.006271111054552927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,512,0.0036151111125946045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,512,0.006109333286682765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,256,0.00462755560874939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,512,0.004986666556861666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,256,0.0032515554792351196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,256,0.0057813334796163775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1536,128,0.004373333520359463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,128,0.002938666691382726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,128,0.005575110928879843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,64,0.0028213332924577924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,64,0.00565155554148886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1536,32,0.0029360000044107437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1536,32,0.005694222119119432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,65536,0.050664888487921826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,65536,0.07271733548906115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,16384,0.015807999504937064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,16384,0.022430222895410325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,12288,0.020113777783181932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,12288,0.013453333742088742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,12288,0.018894222047593858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,10240,0.01717155509524875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,10240,0.012315555579132505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,10240,0.01679733395576477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,8192,0.015293333265516492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,16384,0.024475556280877855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,65536,0.08860444360309178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,8192,0.013047999805874295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,8192,0.014808000789748298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,7168,0.013531555732091268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,7168,0.011580444044537015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,7168,0.01384888920519087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,6144,0.01093777765830358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,6144,0.012888000243239932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,5120,0.010490666660997601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,5120,0.01015377789735794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,5120,0.011223999990357293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,6144,0.011624889241324531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,4096,0.009176889227496253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,4096,0.009314667019579146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,4096,0.009394666386975182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,3584,0.008535111116038429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,3584,0.00907022257645925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,3584,0.008921778036488427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,3072,0.007810666329330868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,3072,0.008423999779754216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,3072,0.0083155557513237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,2560,0.007195555501514011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,2560,0.007961778177155389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,2048,0.006461333483457565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,2560,0.006872889068391588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,2048,0.005955555372767978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,2048,0.0074373334646224976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,1536,0.005973333285914526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,1536,0.005045333256324132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,1536,0.006952889263629913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,1024,0.004323555363549126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,1024,0.006496888895829518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,768,0.005131555514203178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,768,0.003962666624122196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,768,0.006248000181383557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,512,0.004943999979231092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,512,0.003620444486538569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,512,0.005924444645643234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,256,0.0032293332947625052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,256,0.004601777841647466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,256,0.005692444327804778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,1024,0.005424889011515512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,1024,128,0.004366222355100843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,128,0.0029191111938820947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,128,0.005564444594913059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,64,0.0027875554644399216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,64,0.005580444302823808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,1024,32,0.0028515555378463534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,1024,32,0.005560889012283749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,65536,0.07716088824801974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,65536,0.04257155458132426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,65536,0.0692124432987637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,16384,0.021720889541837905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,16384,0.013209777573744455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,16384,0.022250667214393616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,12288,0.01771022213829888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,12288,0.012906666431162091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,12288,0.0184560004207823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,10240,0.015630222029156156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,10240,0.011863999896579318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,10240,0.01676977839734819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,8192,0.013399110900031196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,8192,0.010480889015727572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,8192,0.014448000325096978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,7168,0.012092444631788464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,7168,0.009876444107956356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,7168,0.01295377810796102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,6144,0.010592000352011787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,6144,0.010103111465771994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,6144,0.01185955521133211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,5120,0.009425777528021071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,5120,0.00962222201956643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,5120,0.010556444525718689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,4096,0.008473777936564552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,4096,0.00867555538813273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,4096,0.009195555415418413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,3584,0.008073777788215214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,3584,0.008017777568764156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,3584,0.008787555827034844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,3072,0.0074444446298811175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,3072,0.007291555404663086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,3072,0.008355555435021719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,2560,0.006839111033413146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,2560,0.0063368889192740125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,2560,0.00795733349190818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,2048,0.006380444599522485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,2048,0.005715555614895291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,2048,0.007351111206743453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,1536,0.00591555568906996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,1536,0.004960000101062986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,1536,0.0069191112286514705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,1024,0.0053173332578606075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,1024,0.004253333227501975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,1024,0.006454222318198945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,768,0.005015999906592899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,768,0.003863111138343811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,768,0.006227555374304454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,512,0.004823110997676849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,512,0.003535110917356279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,512,0.005924444645643234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,256,0.003128888913326793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,256,0.005631111148330901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,128,0.004249777644872665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,128,0.00291377781993813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,128,0.005488889084921942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,768,256,0.004545777622196409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,64,0.002793777734041214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,64,0.005550222264395819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,768,32,0.0027795556104845474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,768,32,0.005526222288608551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,65536,0.07380266984303792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,65536,0.03309155503908793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,65536,0.06890933381186591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,16384,0.021507556239763897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,16384,0.012727999852763282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,16384,0.022223111655977037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,12288,0.017581333716710407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,12288,0.010809777511490716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,12288,0.018218666315078735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,10240,0.015344889627562629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,10240,0.010079111489984725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,10240,0.016209777858522203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,8192,0.012488000094890594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,8192,0.009314667019579146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,8192,0.013973333769374423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,7168,0.008775110873911116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,7168,0.012568888564904531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,6144,0.008255999949243333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,6144,0.011167999770906238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,5120,0.009301332963837517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,7168,0.011504000259770287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,5120,0.008169777691364288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,5120,0.010213333699438307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,4096,0.008295111358165741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,4096,0.0074879998962084455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,6144,0.0103502223889033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,4096,0.0090844440791342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,3584,0.007916444705592262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,3584,0.007445333732499017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,3584,0.00872622181971868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,3072,0.007311111523045435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,3072,0.006983111302057902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,3072,0.008246222303973304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,2560,0.006752000086837345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,2560,0.006248888870080312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,2560,0.007696888513035244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,2048,0.0062977779242727495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,2048,0.005614222337802251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,2048,0.007233777807818518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,1536,0.005759999983840519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,1536,0.00490044429898262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,1536,0.006763555523422029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,1024,0.005236444373925527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,1024,0.0041617775956789655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,1024,0.0063591111037466265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,768,0.005024000174469418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,768,0.003802666647566689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,768,0.006138666636413998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,512,0.0047048889100551605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,512,0.0034479999707804788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,512,0.005850666513045629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,256,0.004416888786686791
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,256,0.003077333379122946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,256,0.005569777968857024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,512,128,0.0041884444653987885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,128,0.0028533333291610083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,128,0.00545155546731419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,64,0.002708444371819496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,64,0.005483555710977978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,512,32,0.002740444408522712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,512,32,0.005501333210203383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,65536,0.07354844278759427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,65536,0.02158222264713711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,65536,0.06867111391491361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,16384,0.02128711177243127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,16384,0.009325332939624786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,16384,0.02184266679816776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,12288,0.016706667012638517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,12288,0.008556444611814287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,12288,0.017289777596791584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,10240,0.014371555712487964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,10240,0.008047999607192146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,10240,0.015254222684436373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,8192,0.007378666765160031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,8192,0.012796444197495779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,7168,0.011323555476135679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,8192,0.012234666281276278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,7168,0.007123555574152205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,7168,0.011910222470760345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,6144,0.010261333651012844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,6144,0.006770666688680649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,6144,0.010961777634090848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,5120,0.007079111205206976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,5120,0.010261333651012844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,4096,0.008223111430803934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,4096,0.006737777756320105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,4096,0.009108444054921469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,5120,0.009208888643317753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,3584,0.007135999699433644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,3584,0.00868622213602066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,3584,0.007790221936172909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,3072,0.007237333390447829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,3072,0.006783111227883234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,3072,0.008175111479229396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,2560,0.006633777585294511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,2560,0.006115555349323485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,2560,0.007692444655630324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,2048,0.006197333335876465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,2048,0.005423999908897612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,2048,0.007164444360468123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,1536,0.005728888842794631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,768,0.004927111168702443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,1536,0.0047600000268883174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,1536,0.006688000013430913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,1024,0.00407911092042923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,1024,0.006239999913507038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,768,0.0037271111375755737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,768,0.006049777898523543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,512,0.004564444637960858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,512,0.0034106667670938703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,512,0.005787555542257097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,1024,0.005119110975000593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,256,0.003051555612021022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,256,0.00554933316177792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,128,0.0028311111446883944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,128,0.005388444496525659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,64,0.0026764445420768526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,64,0.005430222385459476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,256,32,0.0026951111439201566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,256,32,0.005437333136796951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,65536,0.07332711087332831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,65536,0.015678221980730694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,65536,0.06863200002246432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,16384,0.02050755586889055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,16384,0.007656888829337225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,16384,0.020986666282018025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,12288,0.00720355576939053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,12288,0.016561778055297006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,10240,0.014242667290899487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,10240,0.0070497774415545994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,10240,0.014501333236694336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,8192,0.01218488895230823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,8192,0.006980444822046492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,8192,0.012320000264379712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,7168,0.011275555524561139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,7168,0.006820444431569841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,7168,0.011663999822404651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,6144,0.006574222197135289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,6144,0.010770666930410596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,5120,0.009191999832789103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,5120,0.0070017774899800616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,256,0.004341333276695675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,5120,0.009909333454238044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,256,128,0.004093333250946469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,12288,0.016467556357383728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,6144,0.010269333091047075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,4096,0.008198222352398766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,4096,0.006584888945023219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,4096,0.008824889030721452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,3584,0.007050666544172499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,3584,0.007711999946170383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,3584,0.008432889150248634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,3072,0.007178666690985362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,3072,0.006726222319735422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,3072,0.008106666306654612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,2560,0.006635555376609166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,2560,0.006117333140638139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,2560,0.007560888926188151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,2048,0.005449777675999536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,2048,0.0070675553547011475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,1536,0.004783111313978831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,1536,0.006726222319735422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,1024,0.004043555508057277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,1024,0.006139555739031897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,768,0.0047600000268883174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,768,0.003719999972316954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,768,0.005978666659858491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,512,0.004583111239804162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,2048,0.006112888869312074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,512,0.0033733333564466904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,1536,0.005622222191757626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,1024,0.005063111169470681
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,512,0.005737777799367905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,256,0.00301155551440186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,256,0.005468444277842839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,128,0.0028168888141711554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,128,0.004101333518822988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,128,0.005411555369695027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,64,0.002672000063790215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,96,128,256,0.004264889078007804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,64,0.005364444520738389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,128,32,0.0026622222115596137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,128,32,0.005335111171007156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,65536,0.013526221944226159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,65536,0.06828177637524076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,16384,0.006687110910813014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,16384,0.019861333900027804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,12288,0.006626666833957036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,12288,0.01586666703224182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,10240,0.006776000062624614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,10240,0.014135999812020196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,8192,0.006664000037643645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,8192,0.012261333564917246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,7168,0.006568000134494569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,7168,0.011482666763994427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,6144,0.0064151109092765385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,6144,0.010661332971519895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,5120,0.006864000111818314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,5120,0.00980533328321245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,4096,0.006528000036875407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,4096,0.008736000292830998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,3584,0.0069511110583941145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,3584,0.008408000071843466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,3072,0.006616000086069107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,3072,0.008057778080304464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,2560,0.006041777630647023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,2560,0.0075093333919843035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,2048,0.005414222263627582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,2048,0.0070497774415545994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,1536,0.0047146665553251905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,1536,0.006675555474228329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,1024,0.004032000071472592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,1024,0.006170666466156642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,768,0.003624000069167879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,768,0.005967111223273807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,512,0.0033662221911880705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,512,0.0057502223385704895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,256,0.002990222225586573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,256,0.005584888988071018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,128,0.0028168888141711554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,128,0.005398222141795688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,64,0.0026195556339290407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,64,0.005400889035728242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,64,32,0.002653333254986339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,64,32,0.005387555393907759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,65536,0.01274400038851632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,65536,0.06844177511003283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,12288,0.006544000158707301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,12288,0.01587022178702884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,10240,0.006722666737106111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,16384,0.019578667150603402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,16384,0.006677333265542984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,10240,0.014214222629865011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,8192,0.006604444649484422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,7168,0.006567111031876669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,8192,0.012375999655988483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,7168,0.011496889094511667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,6144,0.006412444429265128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,6144,0.01071466671095954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,5120,0.006803555621041193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,5120,0.009883555273214975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,4096,0.0087333329849773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,4096,0.0063946665161185795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,3584,0.006870222174459034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,3584,0.00832088871134652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,3072,0.006517333288987477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,3072,0.008078222473462423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,2560,0.006038222048017714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,2560,0.0074924445814556545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,2048,0.005374222166008419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,2048,0.007024889190991719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,1536,0.004713777866628435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,1536,0.006603555546866522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,1024,0.003995555556482739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,1024,0.006214222146405115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,768,0.003629333443111844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,768,0.005931555396980709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,512,0.003264888914095031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,512,0.005704000178310607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,256,0.0029537777105967202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,256,0.005471999860472149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,128,0.002708444371819496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,128,0.005392000079154968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,64,0.002570666579736604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,64,0.0053591111467944244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,96,32,32,0.002575111058023241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,96,32,32,0.00534400012758043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,16384,0.2897964318593343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,16384,0.5671271218193902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,12288,0.42907023429870605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,12288,0.22198132673899332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,16384,0.5562044249640571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,10240,0.3579555617438422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,12288,0.43966489368014866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,10240,0.18030844794379342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,8192,0.28419823116726345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,8192,0.15239200327131483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,10240,0.3665182325575087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,8192,0.2982951005299886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,7168,0.2549164560106066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,7168,0.13425421714782715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,6144,0.1151768896314833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,7168,0.2721039983961317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,6144,0.22156088882022432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,5120,0.09754488865534465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,6144,0.2274115615420871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,5120,0.18453689416249594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,4096,0.0761537750562032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,4096,0.1488000022040473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,5120,0.19457599851820204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,3584,0.06900444295671251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,3584,0.13040799564785427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,4096,0.1577155590057373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,3584,0.14446843994988337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,3072,0.11412533124287923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,3072,0.06091377470228407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,3072,0.12089155779944526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,2560,0.053018665975994535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,2560,0.09535911348130967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,2560,0.10168088807000054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,2048,0.044032000833087497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,2048,0.0811777777141995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,2048,0.08086044258541532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,1536,0.03562755717171563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,1536,0.060425778230031334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,1536,0.061408003171284996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,1024,0.02773422168360816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,1024,0.04336888922585381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,1024,0.043720000320010714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,768,0.034128000338872276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,256,0.01519377695189582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,768,0.03373510970009698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,768,0.023843554986847773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,512,0.018952889574898612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,512,0.02590933276547326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,512,0.025480000509156123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,256,0.014246222045686511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,256,0.021935999393463135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,65536,128,0.011738667057620155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,128,0.009583111438486312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,128,0.02029333346419864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,64,0.007893333004580604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,64,0.021966222259733412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,65536,32,0.006892444358931647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,65536,32,0.025588444537586633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,16384,0.15077955192989773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,16384,0.12863822778066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,16384,0.0864257746272617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,65536,0.5761609077453613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,65536,0.31742578082614475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,65536,0.5646773444281684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,12288,0.06624799966812134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,12288,0.09866844283209907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,12288,0.10977777507570054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,10240,0.08362666765848796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,10240,0.09253066778182983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,8192,0.04662222332424588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,8192,0.07514133056004842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,8192,0.07779022057851155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,7168,0.039635555611716375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,7168,0.05913155608707004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,7168,0.06696266598171659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,6144,0.03520888752407498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,6144,0.051742222574022084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,6144,0.06068355507320828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,5120,0.030144890149434406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,10240,0.05491111013624403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,5120,0.04412177867359585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,5120,0.050127999650107495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,4096,0.02565866708755493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,4096,0.03692266676161025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,4096,0.042635556724336415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,3584,0.023538667294714186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,3584,0.03214577833811442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,3584,0.038951999611324735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,3072,0.02833333280351427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,3072,0.03195377853181627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,3072,0.020862221717834473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,2560,0.018759111563364666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,2560,0.02461599972512987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,2560,0.02717688845263587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,2048,0.020534222324689228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,2048,0.016287111573749118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,2048,0.023432888918452795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,1536,0.013752000199423896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,1536,0.016615110966894362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,1536,0.01849422189924452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,1024,0.011412444214026133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,1024,0.011997333003415002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,1024,0.01497066683239407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,768,0.010240889257854885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,768,0.009829333258999718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,768,0.013369777964221107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,512,0.009159999589125315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,512,0.007762666377756331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,512,0.011727111207114326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,256,0.0068711112770769335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,256,0.00553955551650789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,256,0.008707555631796518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,16384,128,0.005633777628342311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,128,0.0041368889311949415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,128,0.00794400026400884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,64,0.0038568890757030914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,64,0.008988444175985124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,16384,32,0.004000888930426704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,16384,32,0.009716444545321995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,65536,0.2430488798353407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,65536,0.41176001230875653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,16384,0.10930311017566258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,16384,0.06178400251600477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,65536,0.425239986843533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,16384,0.11136621899074978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,12288,0.0481608874268002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,12288,0.0854968892203437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,12288,0.0839608907699585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,10240,0.04172711239920723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,10240,0.07380800114737616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,10240,0.07123200098673503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,8192,0.03726222117741903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,8192,0.054416000843048096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,8192,0.0611706640985277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,7168,0.03153511219554477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,7168,0.04900000161594815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,7168,0.05411377880308363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,6144,0.027816000911924574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,6144,0.0417795545525021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,6144,0.04598755637804667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,5120,0.024449777272012498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,5120,0.03564622335963779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,5120,0.04012711180580987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,4096,0.02073777715365092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,4096,0.030143999391131934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,4096,0.03278933299912347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,3584,0.019114666514926486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,3584,0.02624622152911292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,3584,0.029887109994888306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,3072,0.01740533279048072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,3072,0.023181334137916565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,3072,0.026203556193245783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,2560,0.015824889143308003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,2560,0.020145777199003432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,2560,0.02154133386082119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,2048,0.0140364451540841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,2048,0.017100445098347135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,2048,0.01956977778010898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,1536,0.012227555943859948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,1536,0.013599110974205865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,1536,0.015400888191329109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,1024,0.01035733355416192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,1024,0.010274666878912184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,1024,0.013036444783210754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,768,0.009339555270142024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,768,0.00872622181971868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,768,0.011781333221329583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,512,0.00815466625822915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,512,0.0070088886552386815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,512,0.009840889109505547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,256,0.005792888916201062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,256,0.004556444370084339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,256,0.007835555407736037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,12288,128,0.0052684446175893145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,128,0.00397866674595409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,128,0.007341333561473423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,64,0.003593777616818746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,64,0.007825777762466008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,12288,32,0.0036275556517971885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,12288,32,0.008422222402360704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,65536,0.21520445081922743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,16384,0.056258665190802686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,16384,0.09765066703160603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,65536,0.36323732799953884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,16384,0.10675911108652751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,12288,0.04374577932887607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,10240,0.06647288799285889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,65536,0.3771617677476671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,12288,0.07716355721155803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,12288,0.07711111174689399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,10240,0.06718044148551093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,8192,0.056109335687425405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,8192,0.0540906654463874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,7168,0.02868266569243537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,7168,0.049991998407575816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,10240,0.037815110550986394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,8192,0.031649778286616005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,7168,0.046089778343836464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,6144,0.044686221414142184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,6144,0.042507555749681264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,5120,0.022293332550260756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,5120,0.02973955538537767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,5120,0.03714222378200955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,4096,0.018931556079122756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,4096,0.024696000748210486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,4096,0.030476444297366675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,3584,0.017667555146747164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,3584,0.02203377750184801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,3584,0.027544000082545813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,3072,0.01603555513752831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,3072,0.019540444016456604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,3072,0.023520888553725347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,2560,0.014602666099866232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,6144,0.0251920007997089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,2560,0.017824000782436795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,2560,0.020058666666348774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,2048,0.01294133315483729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,2048,0.01481066644191742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,2048,0.017497777938842773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,1536,0.011218667030334473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,1536,0.011609777808189392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,1536,0.01421955558988783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,1024,0.009906666974226633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,1024,0.009242666264375051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,1024,0.01223288890388277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,768,0.008823999928103553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,768,0.007924444145626491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,768,0.011351111034552256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,512,0.006817777951558431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,512,0.006605333338181178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,512,0.008926221893893348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,256,0.005618666609128316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,256,0.004456000195609199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,256,0.007664889097213745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,10240,128,0.005155555489990446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,128,0.0038702223036024305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,128,0.00720888872941335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,64,0.0034648889882696998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,64,0.007611555357774098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,10240,32,0.003543111185232798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,10240,32,0.008164444731341468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,65536,0.17281333605448404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,65536,0.27390845616658527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,16384,0.04775911238458422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,16384,0.07353778017891778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,65536,0.29028089841206867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,16384,0.07681244611740112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,12288,0.03594933284653558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,12288,0.058079110251532666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,12288,0.06008177995681763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,10240,0.031502223677105375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,10240,0.05033066537645128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,10240,0.04964888758129544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,8192,0.02699377801683214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,8192,0.036109334892696805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,8192,0.04210755560133192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,7168,0.023451555106374953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,7168,0.03194311261177063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,7168,0.037604444556766085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,6144,0.020938666330443487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,6144,0.02801333202256097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,6144,0.0322035551071167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,5120,0.018947554959191214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,5120,0.02442399991883172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,5120,0.027850667635599773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,4096,0.01610577768749661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,4096,0.0206657780541314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,4096,0.02338933282428318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,3584,0.014872888724009195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,3584,0.01835466590192583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,3584,0.021439111895031396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,3072,0.013594666288958656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,3072,0.016410667035314772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,3072,0.01848355597919888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,2560,0.012277333272827996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,2560,0.014150222142537435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,2560,0.016384000579516094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,2048,0.011139555937714048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,2048,0.011763555308183035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,2048,0.014797333214018079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,1536,0.010026666853162978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,1536,0.009720000127951304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,1536,0.012107555237081317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,1024,0.007716444631417592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,1024,0.010335110955768162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,768,0.007479111353556315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,768,0.006800000038411882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,1024,0.00870222184393141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,768,0.008775110873911116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,128,0.004855999930037392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,512,0.00600888869828648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,512,0.005712000032265981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,512,0.0074826669361856245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,8192,256,0.00535999983549118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,256,0.0039013334446483185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,256,0.006793777975771163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,128,0.0035377778112888336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,128,0.006468444648716185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,64,0.0032951111594835916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,64,0.006953777538405524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,8192,32,0.003474666840500302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,8192,32,0.007281777759393056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,65536,0.2685155603620741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,65536,0.18827822473314074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,16384,0.04995200037956238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,16384,0.0654204421573215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,16384,0.07819466458426581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,65536,0.2823573218451606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,12288,0.036357333262761436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,12288,0.04553777641720242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,12288,0.051763554414113365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,10240,0.031077331966824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,10240,0.03937155670589871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,10240,0.05011022090911865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,8192,0.02587911155488756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,8192,0.0321706665886773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,8192,0.04016266597641839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,7168,0.02316266629430983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,7168,0.028411553965674505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,7168,0.032997333341174655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,6144,0.019880889190567862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,6144,0.02474844455718994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,6144,0.030692444907294378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,5120,0.017881777551439073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,5120,0.02180800007449256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,5120,0.026662222213215295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,4096,0.015669332610236276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,4096,0.01852799952030182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,4096,0.02157955533928341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,3584,0.014295111099878946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,3584,0.016386666231685214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,3584,0.020534222324689228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,3072,0.013163555827405719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,3072,0.014535110857751636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,3072,0.017148445049921673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,2560,0.012333333492279053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,2560,0.012468444804350534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,2560,0.015258666541841296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,2048,0.011267555256684622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,2048,0.010584000084135266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,2048,0.014336888988812765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,1536,0.010049777726332346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,1536,0.008912888665994009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,1536,0.011493333511882357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,1024,0.009162666896979014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,1024,0.007082666787836287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,1024,0.010006222460005019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,768,0.00720888872941335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,768,0.006278222219811545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,768,0.00812977800766627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,512,0.006126222097211414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,512,0.004845333182149463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,512,0.007520889242490132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,256,0.003822222352027893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,256,0.006807999892367258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,128,0.003474666840500302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,256,0.005239999956554837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,7168,128,0.004837333328194088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,128,0.006487999939256244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,64,0.0032319999817344877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,64,0.006783111227883234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,7168,32,0.0033111110743549135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,7168,32,0.007082666787836287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,65536,0.2064951128429837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,65536,0.18721955352359346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,65536,0.2537146674262153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,16384,0.04674044582578871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,16384,0.06002044677734375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,16384,0.07265600230958727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,12288,0.03532977898915609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,12288,0.04026755690574646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,12288,0.05089155501789517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,10240,0.03477777706252204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,10240,0.04398311177889506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,8192,0.025590222742822435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,8192,0.02858755654758877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,8192,0.03806311223242018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,7168,0.023071999351183575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,7168,0.025158221522967022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,7168,0.03121244576242235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,6144,0.022247999906539917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,6144,0.028191109498341877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,10240,0.03059911065631443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,6144,0.020676443974177044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,5120,0.018430221411916945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,5120,0.019458666443824768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,5120,0.024511999554104272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,4096,0.016724444097942777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,4096,0.020336000455750358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,4096,0.01570133368174235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,3584,0.014492443866199918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,3584,0.014551111393504672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,3584,0.020424889193640817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,3072,0.013415999710559845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,3072,0.012728888955381183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,3072,0.01794933279355367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,2560,0.012288889123333825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,2560,0.01128533316983117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,2560,0.015099555253982544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,2048,0.010945777926180096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,2048,0.009735999835862054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,2048,0.013267555170589022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,1536,0.009600888523790572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,1536,0.008255110846625434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,1536,0.011553777588738335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,1024,0.007655111451943715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,1024,0.006660444455014334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,1024,0.009204444785912832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,768,0.006070222291681502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,768,0.005975111077229182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,768,0.008187555604510838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,512,0.005601777798599667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,512,0.004325333154863781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,512,0.0074879998962084455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,256,0.00509333320789867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,256,0.0037804444630940757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,256,0.006769777586062749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,6144,128,0.004713777866628435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,128,0.0034506666577524612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,128,0.006430222342411677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,64,0.003237333355678452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,64,0.006709333509206772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,6144,32,0.0032755554550223877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,6144,32,0.006993778049945831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,65536,0.16233866744571263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,65536,0.18309066030714247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,16384,0.046859555774264865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,16384,0.05121511220932007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,16384,0.06473333305782743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,12288,0.03178933262825012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,65536,0.24674312273661295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,12288,0.041166222757763334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,12288,0.048651556173960366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,10240,0.027430221438407898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,10240,0.036053333017561175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,10240,0.04124177826775445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,8192,0.030797332525253296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,8192,0.024471999870406255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,8192,0.03596444593535529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,7168,0.021222222182485793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,7168,0.028039998478359644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,7168,0.030239999294281006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,6144,0.0188400000333786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,6144,0.024919999970330134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,6144,0.025586666332350835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,5120,0.016196444630622864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,5120,0.0170462214284473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,5120,0.022740445203251306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,4096,0.014108444253603617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,4096,0.014558222558763293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,4096,0.01974311139848497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,3584,0.013175110850069257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,3584,0.012702222499582501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,3584,0.017609778377744887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,3072,0.011414222419261932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,3072,0.011945777469211154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,3072,0.015296889675988091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,2560,0.010713777608341642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,2560,0.010021333065297868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,2560,0.014112000664075216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,2048,0.009884444375832876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,2048,0.008774222599135505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,2048,0.012433778080675336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,1536,0.008820444345474243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,1536,0.0075528886583116316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,1536,0.010957332948843638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,1024,0.006671999891599019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,1024,0.006197333335876465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,1024,0.008481777376598781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,768,0.0061919999619325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,768,0.005422222117582957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,768,0.007995555798212687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,512,0.005609777652555042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,512,0.004040889028045866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,512,0.007231111327807109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,256,0.00508977762526936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,256,0.0036195553839206696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,256,0.006759111252095964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,5120,128,0.0047120000753137796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,128,0.0033617777129014335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,128,0.0064479998416370815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,64,0.003116444374124209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,64,0.006660444455014334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,5120,32,0.0031786666562159858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,5120,32,0.006884444504976273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,65536,0.1464195516374376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,65536,0.14505955908033583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,65536,0.1496044397354126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,16384,0.041746666034062706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,16384,0.04355644517474704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,12288,0.03042311138576931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,12288,0.03368799885114034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,12288,0.03243644369973077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,16384,0.04212177793184916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,10240,0.025451555848121643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,10240,0.028712888558705647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,10240,0.026664000418451097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,8192,0.022630222969584998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,8192,0.024843555357721116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,8192,0.023637332850032385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,7168,0.0192586663696501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,7168,0.022613333331214056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,5120,0.016758221719000075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,7168,0.021190222766664293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,6144,0.017666666044129264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,6144,0.02049777739577823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,6144,0.018593778212865193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,5120,0.014696889453464083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,5120,0.018382221460342407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,4096,0.013361777696344586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,4096,0.016103999482260812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,4096,0.014304000470373364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,3584,0.012328888807031842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,3584,0.011537777880827585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,3584,0.013348444468445249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,3072,0.011133333047231039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,3072,0.010434666441546546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,3072,0.012447111308574677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,2560,0.010026666853162978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,2560,0.009101333717505137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,2560,0.01126933346192042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,2048,0.00885688927438524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,2048,0.007975999679830339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,2048,0.01036444471942054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,1536,0.007600000335110559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,1536,0.0069422221018208405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,1536,0.0086986662613021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,1024,0.006118222243256039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,1024,0.0057884446448749965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,1024,0.007349333001507654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,768,0.005685333162546158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,768,0.00462666650613149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,768,0.00686488880051507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,512,0.005327111317051782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,512,0.003919110943873723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,512,0.0064897777305708984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,256,0.004878222114510006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,256,0.0035297779573334586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,256,0.006011555592219035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,4096,128,0.004590222405062782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,128,0.0032151111712058387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,128,0.005789333333571752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,64,0.0030213333666324615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,64,0.006037333359320958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,16384,0.03734577695528666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,4096,32,0.0037288889288902283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,4096,32,0.006073777874310811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,65536,0.12621422608693442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,65536,0.13936266634199354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,16384,0.03682311044798957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,65536,0.13420711623297796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,16384,0.03905333413018121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,12288,0.029375112719006006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,12288,0.03048355711830987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,12288,0.028963555892308552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,10240,0.024067555864651997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,10240,0.025983111725913152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,10240,0.025880888104438782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,8192,0.020639111598332722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,8192,0.022621333599090576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,8192,0.022122666239738464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,7168,0.0195795562532213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,7168,0.020455110404226515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,7168,0.0199991116921107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,6144,0.01573866605758667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,6144,0.0184542222155465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,6144,0.0181697772608863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,5120,0.014641778336630927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,5120,0.016466667254765827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,5120,0.016332444217469957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,4096,0.013006221916940479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,4096,0.014502222339312235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,4096,0.01391999920209249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,3584,0.011951111257076263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,3584,0.010217777556843227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,3584,0.012900444368521372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,3072,0.011232888533009423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,3072,0.009362666971153682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,3072,0.011973333027627734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,2560,0.00964355551534229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,2560,0.00850311087237464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,2560,0.01108977778090371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,2048,0.008645333349704742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,2048,0.007479111353556315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,2048,0.00980088859796524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,1536,0.0069324444565508105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,1536,0.006719111154476802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,1536,0.008020444048775567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,1024,0.006056889063782162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,1024,0.005605333381228977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,512,0.0063795554969045855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,1024,0.007158222297827403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,768,0.005581333405441708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,768,0.004426666845877965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,768,0.006795555353164673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,512,0.005251555393139522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,512,0.00398488880859481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,256,0.004823110997676849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,256,0.003400000019205941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,256,0.005964444329341252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3584,128,0.004578666554556953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,128,0.0031591111587153543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,128,0.005711110929648082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,64,0.002965333354141977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,64,0.005986666513813867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3584,32,0.003051555612021022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,16384,0.03861244519551595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3584,32,0.006058666855096817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,65536,0.11766489346822102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,65536,0.13573243882921007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,16384,0.035487112071779035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,16384,0.03795911206139459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,65536,0.13070044252607557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,12288,0.027650665905740526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,8192,0.02019733356104957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,12288,0.02604355580276913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,12288,0.028425776296191748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,10240,0.023563555545277063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,10240,0.022983110613293115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,10240,0.02442577812406752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,8192,0.02069155540731218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,8192,0.020802666743596394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,7168,0.018135999639829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,7168,0.018050667312410142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,7168,0.01854311095343696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,6144,0.016439111696349252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,6144,0.01625955601533254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,6144,0.01701155636045668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,5120,0.014156444205178155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,5120,0.014564444621404013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,5120,0.01516888870133294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,4096,0.012616000241703458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,4096,0.013025778035322825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,4096,0.013314666847387949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,3584,0.011759111450778114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,3584,0.012146666646003723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,3584,0.012437333663304647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,3072,0.010635555618339114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,3072,0.011336888704035016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,3072,0.011540444360838996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,2560,0.00981955561372969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,2560,0.008147555920812819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,2560,0.010613333019945355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,2048,0.008095111283991072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,2048,0.007323555648326874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,2048,0.009104889300134447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,1536,0.006498666687144174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,1536,0.007752888732486301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,1024,0.006026666611433029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,1024,0.005503111001518037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,1024,0.007056000332037608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,768,0.00553688903649648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,768,0.0042257776690853965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,768,0.006592000110281839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,1536,0.006698666761318843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,512,0.0052560000783867305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,512,0.0037813331517908308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,512,0.006240889016124938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,256,0.004795555439260271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,256,0.0034017778105205963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,256,0.005913777897755305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,3072,128,0.004461333155632019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,128,0.0031395554542541504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,128,0.005657777604129579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,64,0.002967111145456632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,64,0.0058542220956749385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,3072,32,0.003039111072818438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,3072,32,0.005967111223273807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,65536,0.09286755323410034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,65536,0.12334933545854355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,65536,0.11560622851053874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,16384,0.03393244412210252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,16384,0.02847555610868666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,16384,0.03359199894799127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,12288,0.025455999705526564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,12288,0.02299555473857456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,12288,0.025906667113304138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,10240,0.022742221752802532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,10240,0.020301333732075162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,10240,0.022463111413849726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,8192,0.019312888383865356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,8192,0.017639999588330586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,8192,0.01939288940694597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,7168,0.017050666941536795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,7168,0.015990222493807476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,7168,0.0177839994430542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,6144,0.014282666974597506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,6144,0.016201777590645682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,6144,0.015807999504937064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,5120,0.013768889009952545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,5120,0.013172444370057849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,5120,0.014270222849316068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,4096,0.012482666307025485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,4096,0.0120319997270902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,4096,0.012308444413873883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,3584,0.011296889020336999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,3584,0.010983111129866706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,3584,0.011746666497654386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,3072,0.009972444838947719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,3072,0.008819555242856344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,3072,0.010890666809346942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,2560,0.009383111364311641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,2560,0.007967111137178209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,2560,0.009725333087974125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,2048,0.007287999822033777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,2048,0.008447111480765873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,2048,0.007347555624114142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,1536,0.006593777901596493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,1536,0.006292444550328785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,1536,0.0075128889746136135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,1024,0.005767111149099138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,1024,0.004713777866628435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,1024,0.006814222368929122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,768,0.005519111123349931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,768,0.0041982221106688184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,768,0.006488889041874144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,512,0.005101333475775189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,512,0.003760888758632872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,512,0.006130666782458623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,256,0.003383999897374047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,256,0.0058266665372583605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,128,0.004543999830881755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2560,256,0.0047600000268883174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,128,0.003089777711364958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,128,0.005595555735958948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,64,0.002934222213096089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,64,0.005773333211739858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2560,32,0.0029928889125585556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2560,32,0.005873777800136142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,65536,0.11012888616985744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,65536,0.07558044460084704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,16384,0.03461866577466329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,65536,0.09255111217498779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,16384,0.024067555864651997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,16384,0.02786133355564541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,12288,0.02418666746881273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,12288,0.019738665885395475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,12288,0.020807110601001315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,10240,0.021665778425004747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,7168,0.01402844488620758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,10240,0.017447999782032438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,10240,0.01847111185391744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,8192,0.018537777993414137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,8192,0.015178667174445258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,8192,0.01626844373014238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,7168,0.01737955543729994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,7168,0.014979556202888489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,6144,0.014583999911944071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,4096,0.011982222398122152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,6144,0.01276266657643848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,6144,0.013794666363133324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,5120,0.013451555536852943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,5120,0.011775111158688864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,5120,0.012741333080662621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,4096,0.010775999890433418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,4096,0.011217777927716574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,3584,0.010868444210953183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,3584,0.009528888596428765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,3584,0.010523555179437002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,3072,0.009628444082207149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,3072,0.008868444297048781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,3072,0.009319110876984065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,2560,0.008064000142945183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,2560,0.007837333612971837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,2560,0.00849422232972251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,2048,0.007096889118353526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,2048,0.0069395556218094295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,2048,0.007624000310897827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,1536,0.006337777607970768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,1536,0.005943111247486538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,1536,0.007118222614129384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,1024,0.004545777622196409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,1024,0.005670222143332164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,1024,0.006622222148709827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,768,0.0052933332820733385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,768,0.004171555654870139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,768,0.0063591111037466265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,512,0.0038568890757030914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,512,0.006056889063782162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,256,0.004638222356637319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,256,0.0033333332588275275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,256,0.005766222046481238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,128,0.004419555680619346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,128,0.003104888937539525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,128,0.005542222410440445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,64,0.00291377781993813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,64,0.005636444522274866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,2048,512,0.005187555733654234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,2048,32,0.002955555501911375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,2048,32,0.005734222216738596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,65536,0.09528799851735432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,65536,0.06105689207712809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,65536,0.08324622445636325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,16384,0.026016889346970454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,16384,0.02018044392267863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,16384,0.024235554867320593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,12288,0.020417778028382193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,12288,0.01603822244538201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,12288,0.020192889703644645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,10240,0.018911111685964797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,10240,0.01443555619981554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,10240,0.017302221722073026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,8192,0.016857778032620747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,8192,0.012702222499582501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,8192,0.015488889482286243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,7168,0.014936889211336771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,7168,0.011875555747085147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,7168,0.014470222923490735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,6144,0.011036444041464063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,4096,0.009791110952695211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,6144,0.013160000244776407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,5120,0.010394666757848527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,5120,0.011598221957683563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,6144,0.013109332985348172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,5120,0.012348444097571902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,4096,0.010488000180986194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,3584,0.009441778063774109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,3584,0.00998577806684706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,3584,0.0096124443742964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,3072,0.008216888540320927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,4096,0.01072355525361167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,3072,0.009184888667530483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,3072,0.008826666408114964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,2560,0.0074826669361856245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,2560,0.008188444707128737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,2048,0.006631999793979857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,2048,0.0064186664919058485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,2048,0.007594666547245449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,1536,0.005058666484223472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,2560,0.00720177756415473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,1536,0.006220444622966979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,1536,0.007056000332037608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,1024,0.004345777961942884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,1024,0.005555555638339784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,1024,0.00658044425977601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,768,0.005191111316283544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,768,0.003967111309369405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,768,0.0063279999627007385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,512,0.005039999882380168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,512,0.0036151111125946045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,512,0.006024000131421619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,256,0.00464088883664873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,256,0.0032524443748924467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,256,0.0057706667317284485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1536,128,0.004354666504595014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,128,0.002945777856641345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,128,0.005521777603361342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,64,0.0028311111446883944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,64,0.005640888793600931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1536,32,0.002967111145456632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1536,32,0.005657777604129579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,65536,0.09940799739625718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,65536,0.04950755503442553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,65536,0.06886400116814508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,16384,0.015651555524932016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,16384,0.023688889212078516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,16384,0.030327998929553564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,12288,0.023767999476856653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,12288,0.01330044451687071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,12288,0.01904977858066559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,10240,0.012173333101802401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,10240,0.01998044384850396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,10240,0.01701155636045668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,8192,0.018349332941903006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,8192,0.010980444649855295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,8192,0.014707555373509726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,7168,0.016386666231685214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,7168,0.010519111322032081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,7168,0.013593778014183044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,6144,0.009816889133718278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,6144,0.012798222402731577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,5120,0.01108444482088089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,5120,0.009304000271691216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,5120,0.01125600023402108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,4096,0.010354667074150508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,4096,0.008868444297048781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,4096,0.009420444567998251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,3584,0.009468444519572789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,6144,0.013838222457302941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,3584,0.009026666482289633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,3584,0.008964444200197855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,3072,0.008709333009190029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,3072,0.008380444513426887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,3072,0.00834222220712238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,2560,0.007171555525726742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,2560,0.007932444413503012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,2048,0.006777777853939269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,2560,0.007576888634098901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,2048,0.005713777823580637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,2048,0.0074142225914531285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,1536,0.006107555495368109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,1536,0.005005333158704969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,1536,0.0069448889957533936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,1024,0.005524444497293896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,1024,0.004358222087224324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,1024,0.006445333361625671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,768,0.005362666729423735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,768,0.00397866674595409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,768,0.006208000083764394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,512,0.004979555391603046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,512,0.003576888806290097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,512,0.005962666538026597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,256,0.004594666676388847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,256,0.0032248888164758682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,256,0.005640000104904175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,1024,128,0.004261333495378494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,128,0.002946666752298673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,128,0.005530666559934616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,64,0.0027644443843099806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,64,0.005516444643338521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,1024,32,0.002858666703104973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,1024,32,0.005565333283609814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,65536,0.041730665498309664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,65536,0.06925600104861789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,16384,0.02518400053183238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,16384,0.013007111019558378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,16384,0.022270222504933674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,12288,0.011312888728247749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,12288,0.018453333112928603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,10240,0.01740444368786282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,65536,0.09128710958692764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,12288,0.020217777954207528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,10240,0.01055022246307797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,10240,0.016697777642144095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,8192,0.015218666858143277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,8192,0.011129777464601727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,8192,0.014363555444611443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,7168,0.010882666541470421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,7168,0.012994666894276937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,6144,0.011695110963450538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,6144,0.009715555442704095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,6144,0.011653333074516721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,5120,0.00941777808798684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,5120,0.009587555295891231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,5120,0.010430222584141625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,4096,0.008591999610265097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,4096,0.008546666966544257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,4096,0.009203555683294931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,3584,0.008000888758235509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,3584,0.008295999632941352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,3584,0.008811555802822113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,3072,0.008072888685597314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,3072,0.008344888687133789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,2560,0.00683111117945777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,7168,0.013302221894264221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,3072,0.007694222033023834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,2560,0.006416888700591193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,2560,0.007890666524569193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,2048,0.0063742221229606206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,2048,0.005656888915432825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,2048,0.007364444434642792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,1536,0.005872888697518243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,1536,0.004928888960017098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,1536,0.006872889068391588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,1024,0.0053324442770746015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,1024,0.004285333471165763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,1024,0.0064284445510970215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,768,0.005021333280536863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,768,0.003906666818592283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,768,0.006154666758245892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,512,0.004839111119508743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,512,0.0035457776652442086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,512,0.005889777921968036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,256,0.004504888835880491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,256,0.0031759999692440033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,256,0.00563822231358952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,768,128,0.0041653331783082755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,128,0.0029048888633648553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,128,0.005443555613358815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,64,0.0027599999060233435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,64,0.005514666852023866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,768,32,0.002785777880085839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,768,32,0.005500444521506627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,65536,0.07297955618964301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,65536,0.03277333246337043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,65536,0.06843644380569458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,16384,0.02128711177243127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,16384,0.01146488885084788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,16384,0.02215555475817786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,12288,0.017295999659432303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,12288,0.010135110881593492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,12288,0.0181351105372111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,10240,0.014691554837756686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,10240,0.00960622231165568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,10240,0.016379555066426594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,8192,0.01237511138121287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,8192,0.009167999857001835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,8192,0.013628444737858243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,7168,0.011320888996124268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,7168,0.008718222379684448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,7168,0.012318222059143914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,6144,0.010216889282067617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,6144,0.008193777667151557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,6144,0.011050666371981302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,5120,0.009272889130645329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,5120,0.008041777544551426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,5120,0.010194666683673859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,4096,0.007470221983061896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,4096,0.009105777574910058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,3072,0.008314666648705801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,3584,0.007779555188284979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,3584,0.00756177802880605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,3584,0.008745777938101027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,3072,0.007291555404663086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,3072,0.007276444799370236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,2560,0.006286222073766921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,2560,0.007746666669845581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,4096,0.00832088871134652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,2560,0.006691555596060223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,2048,0.006206222292449739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,2048,0.005590222362014983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,2048,0.007222221957312689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,1536,0.0057884446448749965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,1536,0.004888888862397936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,1536,0.00674933319290479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,1024,0.0041457778877682155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,1024,0.006325333482689328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,1024,0.005203555441564984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,768,0.00498844434817632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,768,0.003830222205983268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,768,0.006120000034570694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,512,0.0046640001237392426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,512,0.0034702221552530923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,512,0.005786666853560342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,256,0.004418666578001446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,256,0.0030826667530669104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,256,0.005589333259397083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,512,128,0.00416355538699362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,128,0.0028444443725877335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,128,0.005404444618357553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,64,0.0027146666414207886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,64,0.005456888841258154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,512,32,0.0027671110712819626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,512,32,0.005444444302055571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,65536,0.07265422079298231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,65536,0.02111022174358368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,65536,0.06827466355429755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,16384,0.0206675562593672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,16384,0.009399111072222391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,16384,0.021596444977654353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,12288,0.016343111793200176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,12288,0.009662222531106737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,12288,0.01700444519519806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,10240,0.014286221729384528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,10240,0.0087333329849773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,10240,0.015133332875039844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,8192,0.01218933363755544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,8192,0.007643555601437886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,8192,0.012535110943847232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,7168,0.011211555865075854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,7168,0.007241778075695038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,7168,0.011921777493423887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,6144,0.01012711144155926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,6144,0.00701333334048589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,4096,0.006765333314736684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,6144,0.011022222538789114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,5120,0.009178666604889764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,5120,0.0075288886825243635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,5120,0.010535111029942831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,4096,0.00817599975400501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,4096,0.009088888764381409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,3584,0.007685333490371704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,3584,0.007124444676770105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,3584,0.008743111458089616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,3072,0.007176888485749562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,3072,0.006853333363930385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,3072,0.008249777886602614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,2560,0.006633777585294511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,2560,0.006117333140638139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,2560,0.007631111476156447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,2048,0.006086222413513396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,2048,0.005430222385459476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,2048,0.007120889094140794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,1536,0.005648889061477449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,1536,0.0047644442982143825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,1536,0.006671999891599019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,1024,0.004067555483844546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,1024,0.005180444568395615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,1024,0.006226666685607698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,768,0.004849777867396672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,768,0.003702222059170405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,768,0.005966222120655908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,512,0.004604444321658877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,512,0.0034302222645945023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,512,0.0057653333577844836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,256,0.004293333325121137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,256,0.0030177777840031516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,256,0.005548444473081165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,256,128,0.0041164445380369825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,128,0.0028026666906144884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,128,0.005304000029961268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,64,0.0026853332916895547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,64,0.005395555661784277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,256,32,0.0026506665680143568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,256,32,0.00535999983549118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,65536,0.01516711049609714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,65536,0.06822311215930514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,65536,0.07245866457621257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,16384,0.02016266683737437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,16384,0.008314666648705801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,16384,0.02032888929049174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,12288,0.016220443778567843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,12288,0.00720000018676122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,12288,0.01626755628320906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,10240,0.007114667031500075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,10240,0.014409777190950183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,8192,0.012118221984969245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,8192,0.0070168889231152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,8192,0.012338666452301873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,7168,0.006858666737874349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,10240,0.014191110928853353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,7168,0.011209777659840055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,7168,0.011571555501884885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,6144,0.010073777702119615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,6144,0.006632888896597757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,6144,0.010672888822025724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,5120,0.0090844440791342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,5120,0.007061333292060428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,5120,0.00980977796845966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,4096,0.008106666306654612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,4096,0.00655733338660664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,4096,0.008759111166000366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,3584,0.007689778175618913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,3584,0.0069982219073507525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,3584,0.008461332983440822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,3072,0.007139555282062954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,3072,0.006700444552633498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,3072,0.00794933322403166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,2560,0.006559111177921295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,2560,0.006132444573773279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,2560,0.007532444265153673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,2048,0.005401777724424998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,2048,0.006992888947327931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,1536,0.005583111196756363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,1536,0.0047377778424157035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,1536,0.006559111177921295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,2048,0.006024888820118374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,1024,0.004051555362012652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,1024,0.006167999986145232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,1024,0.005063999858167437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,768,0.0048062221871482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,768,0.003702222059170405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,768,0.00592533333433999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,512,0.004516444272465176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,512,0.0033742222521040174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,512,0.005712888720962737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,256,0.004305777864323722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,256,0.003039999968475766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,256,0.00546488869521353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,80,128,128,0.004067555483844546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,128,0.00281866660548581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,128,0.005314666777849197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,64,0.0026702222724755607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,64,0.0053671110007498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,128,32,0.002716444432735443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,128,32,0.00535733335547977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,65536,0.013302221894264221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,65536,0.06827288866043091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,16384,0.00703999979628457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,16384,0.019482667247454327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,12288,0.0067528887755341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,12288,0.01573600040541755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,10240,0.006858666737874349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,10240,0.014246222045686511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,8192,0.006793777975771163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,8192,0.012249777714411417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,7168,0.006594666590293248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,7168,0.011436444189813403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,6144,0.006436444405052397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,6144,0.010563555690977307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,5120,0.006913777854707506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,5120,0.00979644474056032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,4096,0.006416888700591193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,4096,0.0086666668454806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,3584,0.006903111106819577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,3584,0.008303999900817871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,3072,0.006664000037643645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,3072,0.008033778104517195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,2560,0.006033777776691649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,2560,0.0075191110372543335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,2048,0.005408888889683618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,2048,0.0069617778062820435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,1536,0.004716444346639845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,1536,0.006552000012662675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,1024,0.00398577791121271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,1024,0.006160888820886612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,768,0.003649777836269803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,768,0.005943111247486538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,512,0.0033413333197434745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,512,0.005686222265164058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,256,0.0029848888516426086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,256,0.005472888963090048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,128,0.002781333401799202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,128,0.005333333379692501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,64,0.0026577777332729767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,64,0.005364444520738389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,64,32,0.0026453334010309647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,64,32,0.0053671110007498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,65536,0.012519111235936483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,65536,0.06813955307006836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,16384,0.007111111448870764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,16384,0.019670221540662978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,12288,0.00666933341158761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,12288,0.015775110986497667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,10240,0.006871999965773688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,10240,0.014039999908871122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,8192,0.006609777609507243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,8192,0.012177777787049612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,7168,0.006493333313200209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,7168,0.011346666349305047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,6144,0.006385777973466449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,3584,0.008202667037645975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,6144,0.010557333628336588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,5120,0.006768888897365994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,5120,0.009694221946928237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,4096,0.006342222293217977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,4096,0.008587555752860175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,3584,0.0069448889957533936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,3072,0.006506666541099548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,3072,0.00795911086930169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,2560,0.006074666563007567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,2560,0.007510222494602203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,2048,0.005399111244413588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,2048,0.0069475554757648045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,1536,0.004683555414279302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,1536,0.006601777755551868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,256,0.002944888960984018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,1024,0.00396799999806616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,1024,0.006142222219043308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,768,0.003633777714437909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,768,0.0058746664888328975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,512,0.0033440000067154565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,512,0.005706666658322017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,256,0.005516444643338521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,128,0.0027253333893087176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,128,0.0053280000057485365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,64,0.002584888910253843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,64,0.005306666509972678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,80,32,32,0.0025884444928831523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,80,32,32,0.005316444569163852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,16384,0.5655822224087185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,16384,0.271541330549452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,12288,0.42778754234313965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,16384,0.27805688646104604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,12288,0.21866310967339408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,12288,0.20019200113084581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,10240,0.3573919932047526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,10240,0.18121066358354357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,8192,0.28493955400254994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,8192,0.141903109020657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,10240,0.18439200189378527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,8192,0.1497031052907308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,7168,0.2636408805847168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,7168,0.12643467055426702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,7168,0.13075378206041124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,6144,0.22044889132181802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,6144,0.11069244146347046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,6144,0.10614489184485541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,5120,0.18370132976108125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,5120,0.08952266640133327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,5120,0.09010222223069932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,4096,0.14778933260175917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,4096,0.07805422279569837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,4096,0.07377066877153185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,3584,0.07001511255900066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,3584,0.12984266546037462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,3584,0.06806933217578463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,3072,0.05727466609742907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,3072,0.11338844564225938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,3072,0.05717155668470594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,2560,0.09396888812383015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,2560,0.049272000789642334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,2048,0.07640000184377034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,2048,0.04394222299257914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,1536,0.033173332611719765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,1536,0.06610133250554402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,2560,0.05024800035688612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,2048,0.04134844409094916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,1536,0.03308088911904229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,1024,0.024844444460339014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,1024,0.04152177770932516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,1024,0.02535022298494975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,768,0.02068622244728936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,768,0.032607999112870954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,768,0.022138666775491502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,512,0.017050666941536795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,512,0.023698666029506262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,512,0.018851555056042142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,256,0.013587555951542325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,256,0.016037333342764113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,256,0.013576889203654395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,65536,128,0.010588444769382477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,128,0.00887111077706019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,128,0.014516444669829475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,64,0.007138667007287343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,64,0.01181688904762268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,65536,32,0.00620977787507905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,65536,32,0.011797332929240333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,65536,0.5800862312316895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,65536,0.2759395440419515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,65536,0.3155457708570692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,16384,0.07147377729415894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,16384,0.13317510816786024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,12288,0.054794665839937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,16384,0.08389333221647476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,12288,0.10230222013261582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,10240,0.04718133476045397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,12288,0.0671057767338223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,10240,0.09752533170912002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,10240,0.05372622278001574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,8192,0.03990133272276984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,8192,0.0689893364906311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,8192,0.04507288999027676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,7168,0.03512444429927402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,7168,0.060177776548597545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,7168,0.04224710994296604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,6144,0.03103111187616984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,6144,0.053101334306928843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,6144,0.035234666532940336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,5120,0.02738933265209198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,5120,0.030776888132095337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,5120,0.04496977726618449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,4096,0.0229404436217414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,4096,0.037167999479505755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,4096,0.025545777546034917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,3584,0.021003555920388963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,3584,0.032700445916917585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,3584,0.023287110858493384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,3072,0.01887822151184082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,3072,0.028937776883443195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,3072,0.02092711130777995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,2560,0.016858667135238647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,2560,0.02492444382773505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,2560,0.01850400037235684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,2048,0.014877332581414117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,2048,0.021458667185571458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,2048,0.016154666741689045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,1536,0.012539555629094442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,1536,0.016404444972674053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,1536,0.014174222946166992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,1024,0.010452444354693094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,1024,0.012155555188655853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,1024,0.01200177768866221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,768,0.009388444324334463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,768,0.009552000297440423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,768,0.010967111421955956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,512,0.007800000409285228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,512,0.010025777750545079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,256,0.005903111149867375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,512,0.00833511104186376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,256,0.006098666538794835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,256,0.007506666911972894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,16384,128,0.00498577786816491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,128,0.0041955556306574075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,128,0.007198221981525421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,64,0.0038080000215106537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,64,0.007233777807818518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,16384,32,0.0038177776667806837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,16384,32,0.007307555940416124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,16384,0.05704177750481499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,65536,0.4052328798505995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,16384,0.09966311189863418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,65536,0.20969600147671172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,16384,0.058396445380316846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,65536,0.23214843538072374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,12288,0.04455911119778951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,12288,0.07587466637293498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,12288,0.04543733265664843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,10240,0.04188444548183017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,10240,0.06488888793521456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,10240,0.042540444268120654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,8192,0.03189333279927572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,8192,0.052390221092436046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,8192,0.03346488873163859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,7168,0.028831998507181805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,7168,0.04586755567126804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,7168,0.029823111163245306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,6144,0.025591999292373657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,6144,0.042252444558673434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,6144,0.026557332939571802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,5120,0.022527111901177302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,5120,0.0347653329372406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,5120,0.02347466680738661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,4096,0.019004444281260174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,4096,0.029623998536003962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,4096,0.019863999552196927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,3584,0.01735733283890618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,3584,0.026288888520664636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,3584,0.018182223041852314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,3072,0.015816888875431485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,3072,0.022467555271254644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,3072,0.016602666841612924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,2560,0.014271999398867289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,2560,0.019351111518012155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,2560,0.01498133275243971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,2048,0.012667555775907306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,2048,0.01610577768749661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,2048,0.013384888569513956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,1536,0.01109955542617374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,1536,0.013016888664828407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,1536,0.011914666328165265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,1024,0.009297777381208207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,1024,0.009575111170609793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,1024,0.010557333628336588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,768,0.008451555338170793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,768,0.008328888979223039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,768,0.009754666851626502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,512,0.007125332951545715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,512,0.006940444310506185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,512,0.0074142225914531285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,256,0.005176888985766305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,256,0.004311999926964442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,256,0.006650666809744305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,128,0.0037671112351947357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,128,0.0064151109092765385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,64,0.0034222222036785553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,64,0.006426666759782367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,12288,128,0.004830222162935468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,12288,32,0.0034773333205117118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,12288,32,0.006492444624503453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,65536,0.3389502101474338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,65536,0.1748382250467936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,16384,0.048455999957190625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,16384,0.09476533201005723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,16384,0.05321333474583096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,12288,0.03802400165134006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,12288,0.07211199733946058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,12288,0.044890665345721774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,65536,0.20533955097198486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,10240,0.03329599897066752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,10240,0.05605066484875149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,10240,0.03638844357596503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,8192,0.028140445550282795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,8192,0.045279999574025474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,8192,0.030624889665179785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,7168,0.025424000289705064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,7168,0.03886311252911886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,7168,0.02755644420782725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,6144,0.022121777137120564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,6144,0.034908443689346313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,6144,0.024591111474566992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,5120,0.019592000378502738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,5120,0.030096001095241968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,5120,0.02197511163022783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,4096,0.016864000095261466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,4096,0.024904888537194993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,4096,0.018630221486091614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,3584,0.015619556109110514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,3584,0.022247110803922016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,3584,0.017131555411550734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,3072,0.014305777019924588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,3072,0.01961511042382982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,3072,0.015779554843902588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,2560,0.012903110848532783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,2560,0.016752888758977253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,2560,0.014093332820468478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,2048,0.011261333194043903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,2048,0.014648889501889547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,2048,0.01275466630856196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,1536,0.01000533335738712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,1536,0.011708444191349877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,1536,0.011576889289749993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,1024,0.008802666432327693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,1024,0.00884622252649731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,1024,0.010130667024188572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,768,0.008060444560315874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,768,0.007981333467695449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,768,0.009137777818573846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,512,0.006250666661394968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,512,0.006149333384301927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,512,0.00703111125363244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,256,0.003942222230964237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,256,0.006605333338181178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,256,0.005079999979999331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,10240,128,0.0047893333766195504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,128,0.0034506666577524612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,128,0.00639111093348927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,64,0.0031911111954185697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,64,0.006430222342411677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,10240,32,0.0032702222880389956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,10240,32,0.006487999939256244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,65536,0.14977422025468615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,16384,0.04097155398792691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,65536,0.26932888560824925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,65536,0.18122577667236328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,16384,0.0781537757979499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,16384,0.05067911081843906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,12288,0.03151822090148926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,12288,0.061054223113589816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,12288,0.039777778916888766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,10240,0.02770311137040456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,10240,0.04999555481804741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,10240,0.034394668208228216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,8192,0.023741333021057978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,8192,0.03797422183884515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,6144,0.02892177634769016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,8192,0.0296088887585534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,7168,0.020970667401949566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,7168,0.03353155652681986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,7168,0.02624711063173082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,6144,0.018823999497625563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,6144,0.02348622183005015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,5120,0.016588444511095684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,5120,0.02492444382773505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,5120,0.0209840006298489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,4096,0.014581332604090372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,4096,0.021401777863502502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,4096,0.01774311065673828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,3584,0.013653332988421122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,3584,0.018544000056054857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,3584,0.016386666231685214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,3072,0.012324444121784635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,3072,0.016470222009552848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,3072,0.014855111638704935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,2560,0.011300444602966309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,2560,0.014084445105658637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,2560,0.013827555709415011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,2048,0.010104888843165504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,2048,0.01186222251918581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,2048,0.012346666720178394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,1536,0.00927466650803884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,1536,0.009704888694816166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,1536,0.011154666543006897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,1024,0.007925333248244392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,1024,0.00777688870827357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,1024,0.009688888986905416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,768,0.0064080001579390625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,768,0.00665244460105896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,768,0.007853333320882585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,512,0.0053688887920644544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,512,0.005574222240183089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,512,0.006883555402358373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,256,0.004942222187916438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,256,0.0038542221817705366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,256,0.006542222367392645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,8192,128,0.004657777647177379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,128,0.003559111307064692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,128,0.006335111127959357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,64,0.0032888888898822996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,64,0.006365333166387346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,8192,32,0.003342222215400802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,8192,32,0.0063999998900625445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,65536,0.1385022269354926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,65536,0.2445111009809706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,16384,0.06601422362857394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,16384,0.03755555550257365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,12288,0.029302222861184016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,65536,0.13001333342658147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,16384,0.03760000069936117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,12288,0.052017778158187866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,12288,0.030332446098327637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,10240,0.02674666709370083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,10240,0.045179555813471474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,10240,0.02676533328162299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,8192,0.021928000781271193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,8192,0.03202933404180739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,8192,0.02278222143650055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,7168,0.028459555572933618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,7168,0.02074666652414534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,6144,0.01772533357143402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,6144,0.024791111548741657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,6144,0.019003555178642273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,5120,0.021743999587164983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,7168,0.019710222880045574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,5120,0.0170462214284473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,5120,0.01576622161600325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,4096,0.01833599971400367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,4096,0.014531556102964612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,3584,0.016328889462682936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,3584,0.013606222139464485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,3072,0.011824000212881299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,4096,0.01370133293999566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,3584,0.01294311136007309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,3072,0.01443911095460256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,3072,0.012700444294346703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,2560,0.012771555946932899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,2560,0.011848889291286469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,2048,0.0096142225795322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,2560,0.010843555960390302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,2048,0.010973333484596677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,2048,0.010836444795131683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,1536,0.008820444345474243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,1536,0.009210666848553551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,1536,0.009865778187910715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,1024,0.007635555333561367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,1024,0.007273777491516537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,1024,0.008382221890820397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,768,0.005980444451173146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,768,0.006393777827421824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,768,0.006649777707126405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,512,0.005225777626037598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,512,0.004837333328194088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,512,0.006234666539563074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,256,0.004840888910823398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,256,0.003847111016511917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,256,0.005952888892756567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,7168,128,0.004555555681387584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,128,0.0034773333205117118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,128,0.005643555687533484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,64,0.003248888792263137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,64,0.00573244442542394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,7168,32,0.003351111171974076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,7168,32,0.005762666463851929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,65536,0.22766489452785918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,65536,0.11664711104498969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,65536,0.11552177535163032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,16384,0.03351022137535943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,16384,0.057855996820661754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,16384,0.034428444173600935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,12288,0.025872000389628943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,12288,0.044101334280437894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,10240,0.02274755636850993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,12288,0.02741688821050856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,10240,0.034874667723973594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,10240,0.024400888217820063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,8192,0.02892177634769016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,8192,0.02111111084620158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,7168,0.02509244448608822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,7168,0.019183110859658983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,6144,0.015697777271270752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,6144,0.022180444664425317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,6144,0.017423111531469557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,5120,0.01423111061255137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,5120,0.01953066719902886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,5120,0.015800888339678448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,4096,0.012443555725945367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,4096,0.016759110821617972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,4096,0.013594666288958656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,8192,0.019555555449591745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,7168,0.01789777808719211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,3584,0.014695111248228284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,3584,0.01278222186697854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,3072,0.010886222124099731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,3072,0.01313244468635983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,3072,0.012011555333932241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,2560,0.011584888729784222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,2560,0.011247999966144562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,2048,0.009191110730171204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,2048,0.010004444254769219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,2048,0.010268444816271463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,3584,0.011828444070286222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,1536,0.008446222378147973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,1536,0.008519111408127679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,1536,0.009420444567998251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,1024,0.006695999867386288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,1024,0.006917333437336816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,768,0.0058853332367208265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,768,0.00646311127477222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,512,0.0052586665583981415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,2560,0.010049777726332346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,512,0.0041146667467223275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,512,0.006207110981146495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,256,0.004818666726350784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,256,0.0036524443162812125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,1024,0.0064426664676931165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,768,0.005636444522274866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,256,0.005884444548024072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,6144,128,0.004587555511130227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,128,0.0033111110743549135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,128,0.005679111099905438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,64,0.0031235555393828284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,64,0.005716444303592046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,6144,32,0.0032195556494924757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,6144,32,0.005731555736727185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,65536,0.10296177864074707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,65536,0.17815021673838297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,16384,0.04995377858479818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,16384,0.030340443054835003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,65536,0.10048533148235744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,16384,0.029304888513353135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,12288,0.02292622294690874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,12288,0.04004355602794223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,8192,0.02959644463327196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,12288,0.024772443705134924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,10240,0.02036977807680766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,10240,0.034797334008746676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,10240,0.022110222114457026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,8192,0.018439999885029264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,8192,0.019150222341219585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,7168,0.015789333317014907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,7168,0.026882666680547927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,7168,0.017303110824690927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,6144,0.024293333292007446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,6144,0.015838222371207345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,5120,0.013001778059535556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,5120,0.01685422162214915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,5120,0.014541332920392355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,4096,0.011537777880827585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,4096,0.014255111416180929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,4096,0.012485333614879184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,3584,0.012520888613329994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,3584,0.011900444825490316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,3072,0.010306666294733683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,3072,0.01108888867828581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,3072,0.011175110936164856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,6144,0.014318222800890604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,3584,0.01105955574247572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,2560,0.009594666461149851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,2560,0.009860444400045607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,2560,0.010533333652549319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,2048,0.008817777865462834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,2048,0.008671110702885522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,2048,0.009703999592198266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,1536,0.00739911115831799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,1536,0.008959111240175035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,1024,0.006105777704053455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,1024,0.0064533332155810455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,1536,0.00814222213294771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,1024,0.00620266670982043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,768,0.005379555539952383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,768,0.006250666661394968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,512,0.00507377791735861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,512,0.003913777983850903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,512,0.006028444402747684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,256,0.004736888739797804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,256,0.0034515555534097883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,768,0.005381333331267039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,256,0.0057813334796163775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,5120,128,0.004507555729813046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,128,0.0032133333798911837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,128,0.005608888963858287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,64,0.00300444434914324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,64,0.005656888915432825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,5120,32,0.0030657777355776895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,5120,32,0.00566133318675889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,65536,0.0867991116311815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,65536,0.14180889394548205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,65536,0.08934666713078816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,16384,0.025095999240875244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,16384,0.04130311144722833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,16384,0.02718399961789449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,12288,0.019670221540662978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,12288,0.032877332634396024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,12288,0.022278222772810195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,10240,0.01757066614098019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,10240,0.028513779242833454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,10240,0.01995022263791826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,8192,0.015587555037604438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,8192,0.02443733314673106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,8192,0.017127111554145813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,7168,0.014177777700954013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,7168,0.022044445077578228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,7168,0.015835555063353646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,6144,0.012991999586423239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,6144,0.019744889603720773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,6144,0.01457244488928053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,5120,0.011818666425016193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,5120,0.01759733259677887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,5120,0.01346933344999949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,4096,0.010757333702511258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,4096,0.012562666502263812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,4096,0.01163200040658315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,3584,0.010218666659461128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,3584,0.010849778023031024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,3584,0.01107911103301578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,3072,0.009553777674833933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,3072,0.009863110880057016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,3072,0.01055555542310079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,2560,0.008967999782827165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,2560,0.008791111409664154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,2560,0.00998222248421775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,2048,0.00831288927131229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,2048,0.007711999946170383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,2048,0.009189333352777693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,1536,0.007124444676770105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,1536,0.006772444479995304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,1536,0.007432888779375289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,1024,0.005517333332035277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,1024,0.005704888867007361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,1024,0.006303999986913469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,768,0.004359111189842224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,768,0.006109333286682765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,512,0.005122666557629903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,768,0.0053084443012873335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,512,0.00378933341966735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,512,0.005904888941182031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,256,0.0047280001971456735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,256,0.003350222276316749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,256,0.00563466673096021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,4096,128,0.004451555510361989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,128,0.003096888876623578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,128,0.005480000128348668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,64,0.0029404444826973807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,64,0.005528888768619961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,4096,32,0.0031866667171319327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,4096,32,0.005610666755172942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,65536,0.12244266933865017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,65536,0.07815644476148817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,65536,0.08163822359508939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,16384,0.02383288906680213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,16384,0.0359555549091763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,16384,0.025993777645958796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,12288,0.018967999352349173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,12288,0.02850666642189026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,12288,0.022230222821235657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,10240,0.016960889101028442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,10240,0.025192888246642217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,10240,0.019194665882322524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,8192,0.014952000644471912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,8192,0.021779555413458083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,8192,0.01647911138004727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,7168,0.013608000344700284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,7168,0.01962933275434706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,7168,0.01531644496652815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,6144,0.012727110750145383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,4096,0.011326221956147088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,6144,0.017650667164060805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,6144,0.014314666390419006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,5120,0.011495999991893768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,5120,0.015777778294351365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,5120,0.013082666529549493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,4096,0.010438222024175854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,4096,0.011396444506115384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,3584,0.009964444571071202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,3584,0.010121777653694153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,3584,0.011029333704047732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,3072,0.00923911150958803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,3072,0.010317333042621613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,2560,0.008330666356616551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,2560,0.009703111317422655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,2048,0.008045333127180735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,3072,0.009486222432719337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,2560,0.008784888519181145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,2048,0.007407111426194509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,2048,0.00850399997499254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,1536,0.006312888943486744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,1536,0.006653333289755716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,1536,0.0070266665683852295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,1024,0.005481777919663324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,1024,0.005326222214433882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,1024,0.006329777754015393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,768,0.005225777626037598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,768,0.0042160000238153665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,768,0.006096000058783426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,512,0.003744889050722122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,512,0.005919111271699269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,256,0.004609777695602841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,512,0.005060444275538127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,256,0.0033564445459180405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,256,0.005696889013051987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3584,128,0.004380444271696938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,128,0.00309599998096625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,128,0.005497777627574072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,64,0.002966222249799304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,64,0.005519111123349931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3584,32,0.0030097777230872046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3584,32,0.00554577757914861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,65536,0.07722400294409858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,65536,0.10713777939478557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,65536,0.07807111077838473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,16384,0.023500444160567388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,16384,0.03239911132388645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,16384,0.025199999411900837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,12288,0.018765333626005385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,12288,0.025633777181307476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,12288,0.020672889219390023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,10240,0.01676177812947167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,10240,0.022602667411168415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,10240,0.018591110905011494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,8192,0.0199617776605818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,8192,0.01607822212908003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,7168,0.013349333571063148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,7168,0.017975111802419026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,7168,0.014863999353514777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,6144,0.01238222254647149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,6144,0.01610666679011451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,6144,0.013759111364682516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,8192,0.014336888988812765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,5120,0.014392889208263822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,5120,0.011306666665607028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,5120,0.01279022213485506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,4096,0.010142222046852112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,4096,0.010583110981517367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,4096,0.011165333290894827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,3584,0.009700444837411245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,3584,0.009566222627957663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,3584,0.010763555765151978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,3072,0.009181333084901174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,3072,0.00886044485701455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,3072,0.010140444669458602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,2560,0.008563555777072906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,2560,0.007986666427718269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,2560,0.009425777528021071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,2048,0.007045333584149678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,2048,0.00720266666677263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,2048,0.007708444363541073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,1536,0.006383111079533895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,1536,0.00685777763525645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,1024,0.005408888889683618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,1024,0.004881777697139316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,1024,0.006320000108745363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,768,0.005191111316283544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,768,0.004156444635656145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,768,0.006106666806671355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,512,0.004860444615284602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,512,0.0037137779096762338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,1536,0.006009777800904379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,512,0.005919111271699269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,256,0.004583999928500918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,256,0.0033368888414568375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,256,0.005636444522274866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,3072,128,0.004330666528807746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,128,0.003030222323205736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,128,0.005476444545719359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,64,0.0028613333900769553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,64,0.005525333185990651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,3072,32,0.002974222310715251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,3072,32,0.005552000055710475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,65536,0.09099200036790635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,65536,0.07681688997480604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,16384,0.023181334137916565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,65536,0.0741928882069058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,16384,0.028417779339684382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,16384,0.024290665984153748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,12288,0.018402665853500366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,12288,0.022571555442280237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,12288,0.019897777173254225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,10240,0.016494222813182406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,10240,0.019871110717455547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,10240,0.017747556169827778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,8192,0.014287999934620328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,8192,0.017247110605239868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,8192,0.015469332536061605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,7168,0.013176889055305056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,7168,0.01570933394961887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,7168,0.014466666513019137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,6144,0.014165333575672574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,6144,0.013458666702111563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,5120,0.012856889102194043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,5120,0.012473777764373355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,4096,0.010068444742096795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,4096,0.011721777419249216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,4096,0.010852444503042432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,3584,0.00960355583164427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,3584,0.009331555830107795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,3584,0.010449777874681683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,6144,0.012123555772834353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,5120,0.011159111228254108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,3072,0.008923555413881937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,3072,0.008538666698667739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,3072,0.009849777652157677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,2560,0.008150222400824228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,2560,0.007808888951937358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,2560,0.008399999803966945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,2048,0.006424888968467712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,2048,0.007113777928882175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,2048,0.007275555696752336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,1536,0.005976888868543837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,1536,0.006100444330109491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,1536,0.006659555352396435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,1024,0.005389333185222413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,1024,0.00442844463719262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,1024,0.006259555617968242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,768,0.0041235557032956015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,768,0.006063111126422882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,512,0.003689777933888965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,512,0.005864000154866113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,256,0.004493333399295807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,256,0.0033288889875014624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,256,0.00562755556570159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,128,0.004282666577233208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,128,0.005470222069157495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,128,0.003039999968475766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,768,0.005128889034191768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,64,0.0029351111087534162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,64,0.005488889084921942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2560,512,0.004874666531880696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2560,32,0.0029964444951878656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2560,32,0.005511111269394557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,65536,0.07282844516966078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,65536,0.07524800300598145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,16384,0.021909332937664453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,65536,0.07141600052515666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,16384,0.023193778263198003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,16384,0.023408888114823237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,12288,0.017680888374646504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,12288,0.019208888212839764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,12288,0.01907199952337477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,10240,0.016920000314712524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,10240,0.017315554949972365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,8192,0.014890667464998035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,10240,0.015643555257055495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,8192,0.014999111493428549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,7168,0.012711111042234631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,7168,0.013906665974193148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,7168,0.013941332697868347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,6144,0.011665778027640449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,6144,0.012688889271683164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,8192,0.013638222383128272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,6144,0.013074666261672974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,5120,0.010667555862002902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,5120,0.011786667009194693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,5120,0.012207110722859701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,4096,0.00961688905954361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,4096,0.010584000084135266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,4096,0.010585777461528778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,3584,0.009111999637550777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,3584,0.009871111147933537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,3584,0.009905777871608734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,3072,0.00795288880666097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,3072,0.009174221919642555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,3072,0.008651555412345463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,2560,0.00700800038046307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,2560,0.0075502221783002215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,2048,0.0063484443558586975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,2560,0.007560888926188151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,2048,0.006816888848940532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,1536,0.005111111121045219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,2048,0.006955555743641323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,1536,0.006575999988449945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,1536,0.005848000033034219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,1024,0.0052915554907586836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,1024,0.004344888859324985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,768,0.00397866674595409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,1024,0.006144888699054718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,768,0.005983110931184556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,512,0.003621333175235324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,512,0.004849777867396672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,256,0.005501333210203383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,512,0.005738666488064661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,256,0.0032622222271230486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,256,0.004530666602982415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,128,0.00434666665063964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,128,0.0029599999801980127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,128,0.005423111220200856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,64,0.0028159999185138275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,2048,768,0.005103999955786599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,64,0.005399111244413588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,2048,32,0.002903999967707528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,2048,32,0.005372444374693765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,65536,0.0611413319905599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,65536,0.06960444317923652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,16384,0.01983644399378035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,16384,0.022639110684394836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,12288,0.01605333387851715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,12288,0.018499554859267343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,10240,0.015267555912335714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,10240,0.014284445179833306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,10240,0.01682044400109185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,8192,0.01331555595000585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,8192,0.012714666624863943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,8192,0.014557333456145393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,7168,0.012455110748608908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,65536,0.07227644655439588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,16384,0.0214720004134708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,12288,0.017264889346228707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,7168,0.011740444435013665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,7168,0.013669333524174161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,6144,0.01146755533085929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,6144,0.010975999964608086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,6144,0.012777778009573618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,5120,0.010434666441546546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,5120,0.010355555348926121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,5120,0.011891555454995898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,4096,0.008822222550710043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,4096,0.009552000297440423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,4096,0.009528000321653154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,3584,0.007954667011896769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,3584,0.00923999978436364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,3584,0.008623110751310984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,3072,0.008780444661776224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,3072,0.007872000336647034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,2560,0.006741333338949416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,2560,0.00703999979628457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,2560,0.007511110769377814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,2048,0.006271111054552927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,3072,0.007312888900438945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,2048,0.006001777946949005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,2048,0.007015110717879401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,1536,0.005783111270931032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,1536,0.004964444372389051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,1536,0.006599111275540457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,1024,0.005221333354711533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,1024,0.004286222159862518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,1024,0.006186666587988536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,768,0.005021333280536863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,768,0.003925333420435588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,768,0.005997333261701796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,512,0.004750222381618288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,512,0.003559999995761447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,512,0.0057839999596277876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,256,0.004371555729044808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,256,0.0031662223239739737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,256,0.005584888988071018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1536,128,0.0041928887367248535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,128,0.0029271110478374693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,128,0.00543200017677413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,64,0.0027644443843099806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,64,0.005469333380460739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1536,32,0.0028435554769304064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1536,32,0.005431111074156231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,65536,0.04383822282155355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,65536,0.06829244560665555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,16384,0.021290666527218286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,16384,0.015601777368121676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,16384,0.022037333912319605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,12288,0.017112000121010672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,12288,0.013160000244776407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,12288,0.018132444885041978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,10240,0.0151573336786694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,10240,0.01221777747074763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,10240,0.016466667254765827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,8192,0.013039111263222165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,8192,0.010679110884666443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,6144,0.009580444130632613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,8192,0.014211555322011312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,7168,0.012030222349696688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,7168,0.01014311114947001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,65536,0.07218844360775419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,6144,0.01054133309258355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,7168,0.013334222137928009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,6144,0.0119982221060329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,5120,0.009304000271691216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,5120,0.009168888959619734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,5120,0.010318222145239512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,4096,0.008135110967689091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,4096,0.008627555436558193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,4096,0.008815111385451423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,3584,0.007654222349325816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,3584,0.008520888785521189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,3584,0.008303999900817871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,3072,0.007800000409285228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,3072,0.007801777786678738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,3072,0.007170666423108842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,2560,0.006620444357395172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,2560,0.006913777854707506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,2560,0.007460444337791867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,2048,0.006116444451941385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,2048,0.0057751110030545135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,2048,0.0069448889957533936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,1536,0.00490399988161193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,1536,0.006554666492674086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,1536,0.00563733321097162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,1024,0.005120888766315249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,1024,0.004257777912749184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,1024,0.006184000107977126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,768,0.004903111192915174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,768,0.00397155558069547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,768,0.005963555640644497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,512,0.0047253333032131195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,512,0.005712000032265981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,512,0.003607999947335985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,256,0.004330666528807746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,256,0.0032524443748924467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,256,0.005519999812046687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,1024,128,0.0041582224269707995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,128,0.002942222274012036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,128,0.005377777748637729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,64,0.002777777819169892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,64,0.005434666656785541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,1024,32,0.0029128889242808023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,1024,32,0.005420444326268301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,65536,0.07105866405698988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,65536,0.03617599937650893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,65536,0.06803199979994032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,16384,0.021027555068333943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,16384,0.012729778057999082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,16384,0.021956443786621094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,12288,0.011190222369299995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,12288,0.018079999420377944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,10240,0.015114666687117683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,10240,0.010245333115259806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,10240,0.01647555496957567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,8192,0.012964444855848948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,8192,0.009345778160625035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,8192,0.013971555564138623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,7168,0.011368888947698804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,7168,0.009001777403884465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,7168,0.012221333053376941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,6144,0.01016088906261656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,6144,0.008597333398130205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,6144,0.010866666833559671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,5120,0.00906488878859414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,5120,0.00870577742656072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,5120,0.009926222264766693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,4096,0.008064889245563084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,4096,0.007496889266702864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,4096,0.008657777474986183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,3584,0.007644444704055786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,3584,0.007637333538797166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,12288,0.016992888516849942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,3584,0.008250666989220513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,3072,0.007123555574152205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,3072,0.006963555183675554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,3072,0.007894222107198503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,2560,0.006647111227114995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,2560,0.007466666400432587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,2048,0.005583999885453119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,2048,0.0069777775141927935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,1536,0.005616000129116907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,1536,0.00488711107108328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,1536,0.006579555571079254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,2560,0.006590222318967183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,1024,0.005103111267089844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,1024,0.0041911109454101985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,1024,0.006194666855865055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,2048,0.006062222023804982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,768,0.004825777891609403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,768,0.003814222084151374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,768,0.006022222340106964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,512,0.0046737777690092725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,512,0.003439999909864532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,512,0.005792888916201062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,256,0.004284444368547863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,256,0.003060444361633725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,256,0.005573333137565189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,768,128,0.004129777765936322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,128,0.002850666642189026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,128,0.005430222385459476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,64,0.002736888825893402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,64,0.005446222093370225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,768,32,0.002814222127199173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,768,32,0.005416000054942236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,65536,0.07185066408581205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,65536,0.027082666754722595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,65536,0.06750222047170003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,16384,0.021003555920388963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,16384,0.011007111105653973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,16384,0.021606221795082092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,12288,0.016906667086813185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,12288,0.00982488857375251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,12288,0.017882666654056974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,10240,0.014470222923490735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,10240,0.009323555562231276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,10240,0.015450666348139444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,8192,0.011922666596041786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,8192,0.008809777597586313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,8192,0.012713777522246042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,7168,0.00833333366447025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,7168,0.011528888510333167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,6144,0.010089777410030365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,6144,0.007423111134105259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,6144,0.010471111370457543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,5120,0.00907022257645925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,5120,0.007663110891977946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,5120,0.009708444277445475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,4096,0.008076444268226624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,4096,0.0069102222720781965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,4096,0.008578666382365758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,7168,0.011011555790901184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,3584,0.007184888753626082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,3584,0.008240000241332585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,3072,0.007066667079925537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,3584,0.007640000018808577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,3072,0.00683822234471639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,3072,0.007810666329330868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,2560,0.006542222367392645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,2560,0.006156444549560547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,2560,0.00739822205570009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,2048,0.006058666855096817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,2048,0.005478222337034013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,2048,0.006950221955776215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,1536,0.005617777920431561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,1536,0.0047733332547876565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,1536,0.006545777950021956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,1024,0.005075555708673266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,1024,0.004091555459631814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,1024,0.006187555690606435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,768,0.00479288895924886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,768,0.003703111161788305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,768,0.005991999887757831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,512,0.004545777622196409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,512,0.0034106667670938703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,512,0.005740444279379315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,256,0.004234666625658671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,256,0.003061333257291052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,256,0.00555288874440723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,128,0.0028684443483750024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,128,0.005432888865470886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,65536,0.07161244418885973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,64,0.0026897777699761917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,64,0.005375999957323074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,512,32,0.0027306667632526825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,512,32,0.005391110976537068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,512,128,0.004122666600677702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,65536,0.01937333411640591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,16384,0.020115555988417733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,65536,0.06705244382222493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,16384,0.008816888762844933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,16384,0.020711110697852243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,12288,0.015979554918077257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,12288,0.008082666330867344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,12288,0.016248888439602323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,10240,0.01402844488620758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,10240,0.007684444387753804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,10240,0.014142221874660917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,8192,0.007466666400432587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,8192,0.012006222373909421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,7168,0.010996444357766045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,7168,0.006872889068391588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,7168,0.011320888996124268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,6144,0.010049777726332346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,8192,0.011980444192886353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,6144,0.006754666566848755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,6144,0.010440000229411654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,5120,0.00904622260067198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,5120,0.007096889118353526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,5120,0.009661333428488838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,4096,0.008020444048775567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,4096,0.006642666541867786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,4096,0.008544888761308458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,3584,0.007575111256705389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,3584,0.007113777928882175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,3584,0.008126222425036961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,3072,0.007061333292060428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,3072,0.006793777975771163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,3072,0.007824888659848107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,2560,0.006116444451941385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,2560,0.007382222347789341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,2048,0.005427555491526921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,2048,0.0069360000391801195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,1536,0.005568000177542369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,1536,0.004775111046102312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,1536,0.006495999793211619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,2560,0.006491555521885554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,2048,0.006004444426960415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,1024,0.00407288885778851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,1024,0.005015999906592899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,1024,0.006099555641412735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,768,0.004824000100294749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,768,0.0036968886852264404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,768,0.005893333504597346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,512,0.004514666481150521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,512,0.0033644443998734155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,512,0.005685333162546158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,256,0.0042160000238153665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,256,0.00305244450767835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,256,0.005490666462315454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,256,128,0.004053333153327306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,128,0.002814222127199173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,128,0.005366222312053044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,64,0.00267999991774559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,64,0.005390222287840313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,256,32,0.00534400012758043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,256,32,0.0027093332674768236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,65536,0.015019555886586508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,65536,0.07139377461539374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,65536,0.06685600015852186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,16384,0.019872888922691345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,16384,0.00720711135201984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,16384,0.01958044370015462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,12288,0.016015999846988253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,12288,0.006680888848172293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,12288,0.015475556254386902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,10240,0.013995556367768182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,10240,0.006867555694447623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,10240,0.013863110707865821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,8192,0.006768888897365994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,8192,0.01201866649919086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,7168,0.010990222295125326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,7168,0.006682666639486949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,7168,0.01128266668981976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,6144,0.009941333697901832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,6144,0.006488889041874144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,6144,0.010410666465759277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,8192,0.011876444849703046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,5120,0.00685066646999783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,5120,0.009651555783218807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,4096,0.006437333507670297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,5120,0.00903200027015474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,4096,0.00853777759604984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,3584,0.0068764446510208984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,3584,0.00812533332241906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,3072,0.0069848886794514125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,3072,0.006568000134494569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,3072,0.00777511133088006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,2560,0.006430222342411677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,4096,0.007984889050324758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,3584,0.0075146663520071245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,2560,0.006072000082996156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,2560,0.007339555356237624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,2048,0.005978666659858491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,2048,0.005387555393907759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,2048,0.0068853331936730286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,1536,0.005481777919663324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,1536,0.0047404443224271136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,768,0.0037288889288902283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,1536,0.00646311127477222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,1024,0.004042666819360522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,1024,0.006064888917737537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,768,0.0047084444926844705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,768,0.0058906666106647914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,512,0.0033813334173626373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,512,0.004439110971159405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,512,0.005662222289376789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,1024,0.00500177757607566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,256,0.0041946665280395085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,256,0.002973333415057924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,256,0.005469333380460739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,64,128,128,0.004011555678314633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,128,0.0028453332682450614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,128,0.005335999859703912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,64,0.0026542221506436667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,64,0.005356444252861871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,128,32,0.0026737778551048706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,128,32,0.005320000151793162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,65536,0.013360888593726687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,65536,0.066721777121226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,16384,0.006694222076071634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,16384,0.019097778532240126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,12288,0.006554666492674086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,12288,0.015451555450757345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,10240,0.006728888799746831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,10240,0.01391999920209249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,8192,0.006653333289755716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,8192,0.012049777640236748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,7168,0.00647555540005366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,7168,0.01131377783086565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,6144,0.006324444380071428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,6144,0.010438222024175854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,5120,0.006670222100284364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,5120,0.009670221971140968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,4096,0.00629333323902554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,4096,0.008589333130253686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,3584,0.006736889067623351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,3584,0.008221333225568136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,3072,0.0064071110553211635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,3072,0.007779555188284979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,2560,0.006031999985376994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,2560,0.007375111182530721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,2048,0.005373333477311664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,2048,0.006919999917348226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,1536,0.005022222383154763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,1536,0.006551110910044775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,1024,0.003986666599909465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,1024,0.006072888771692912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,768,0.003640888879696528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,768,0.005940444353553984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,512,0.0033031110134389666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,512,0.005681777579916849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,256,0.0029493332323100832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,256,0.005500444521506627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,128,0.0027822222974565295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,128,0.0053582220441765254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,64,0.0026497778793176017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,64,0.005315555466545953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,64,32,0.0026542221506436667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,64,32,0.0052791109515561005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,65536,0.012871999707486896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,65536,0.06703733073340522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,16384,0.006552888701359431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,12288,0.0064693333374129395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,16384,0.018904889623324077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,10240,0.006701333241330252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,12288,0.015446222490734525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,10240,0.0138586668504609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,8192,0.006574222197135289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,8192,0.012059555285506778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,7168,0.006380444599522485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,7168,0.011227555572986603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,6144,0.006252444452709622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,6144,0.010395555860466426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,5120,0.006559999866618051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,5120,0.009641778137948778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,4096,0.006182222316662471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,4096,0.00852711084816191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,3584,0.006607999818192587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,3584,0.008157333566082848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,3072,0.006295111030340195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,3072,0.007748444047239091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,2560,0.005978666659858491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,2560,0.007319110963079665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,2048,0.005330666485759947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,2048,0.006893333461549547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,1536,0.004618666652176115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,1536,0.006463999963468975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,1024,0.003946666502290302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,1024,0.006057777752478917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,768,0.0035546666218174826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,768,0.005864000154866113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,512,0.0032915555768542816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,512,0.005648889061477449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,256,0.0029048888633648553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,256,0.005397333453098933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,128,0.00272444449365139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,128,0.0052879999081293745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,64,0.0025191110455327565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,64,0.0052800000541739995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,64,32,32,0.0025680000997251934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,64,32,32,0.005294222384691238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,16384,0.5296604368421767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,12288,0.3965635564592149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,12288,0.21548799673716226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,16384,0.2698711024390327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,10240,0.18164266480339897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,12288,0.22108088599310982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,16384,0.28116001023186576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,10240,0.33121244112650555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,8192,0.26693688498602974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,8192,0.14606666564941406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,10240,0.18753511375851104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,8192,0.15071377489301893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,7168,0.11934222115410699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,7168,0.2337848875257704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,6144,0.201544894112481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,7168,0.12347911463843451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,6144,0.11194399992624919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,5120,0.16917155848609078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,5120,0.08823466963238186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,6144,0.10972088575363159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,4096,0.13719111018710667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,5120,0.09274844328562419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,4096,0.07712977462344699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,4096,0.07716088824801974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,3584,0.06437777810626559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,3584,0.11903644932640924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,3584,0.06633333365122478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,3072,0.10449066427018906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,3072,0.057512885994381376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,3072,0.05716977516810099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,2560,0.08700888686709934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,2560,0.04864266514778137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,2560,0.04971555537647671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,2048,0.040748443868425153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,2048,0.07489333550135295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,1536,0.03253422180811564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,2048,0.04416977696948581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,1536,0.05472177929348416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,1536,0.033088889386918806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,1024,0.038958221673965454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,1024,0.025518221987618342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,1024,0.025360888904995386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,768,0.020319110817379422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,768,0.030498666895760432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,768,0.021739555729760066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,512,0.01666755477587382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,512,0.02218755582968394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,512,0.018820444742838543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,256,0.012871111432711283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,256,0.012605333493815528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,256,0.016006223029560514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,128,0.00831911133395301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,128,0.014468444718254937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,65536,128,0.010098666780524785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,64,0.006779555645253923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,64,0.011550222006109027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,65536,32,0.005568888866239124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,65536,32,0.011672000090281168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,16384,0.12538933753967285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,65536,0.5392933421664768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,65536,0.2745075490739611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,65536,0.32016176647610134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,12288,0.05440444416469998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,12288,0.09593688779407078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,16384,0.08119110928641425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,12288,0.06249244345559014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,10240,0.05030133326848348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,16384,0.0703528920809428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,10240,0.08159555329216851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,10240,0.0578133331404792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,8192,0.039303110705481634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,8192,0.06601333618164062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,8192,0.04470844401253594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,7168,0.034605334202448525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,7168,0.06236622068617079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,7168,0.03975466556019253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,6144,0.030704001585642498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,6144,0.05047288868162367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,6144,0.0351235568523407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,5120,0.02693333393997616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,5120,0.042945779032177396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,5120,0.030741333961486816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,4096,0.022708444131745234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,4096,0.035451554589801364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,4096,0.025693333811230127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,3584,0.020946666598320007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,3584,0.03121333320935567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,3584,0.023563555545277063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,3072,0.01864355636967553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,3072,0.027521777484152053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,3072,0.021041777398851182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,2560,0.016523554921150208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,2560,0.023775999744733173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,2560,0.018573333819707234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,2048,0.019872888922691345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,2048,0.016083555089102853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,1536,0.0159599996275372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,1536,0.014102222190962898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,1024,0.010299555957317352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,1024,0.011618666350841522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,2048,0.0144533332851198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,1024,0.012016889121797351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,768,0.00925155563486947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,768,0.009472889204819998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,1536,0.012429333395428128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,768,0.010832889212502373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,512,0.008212444682916006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,512,0.007429333196745978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,512,0.00980266680320104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,256,0.005543111099137201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,256,0.005513777749405966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,256,0.007430222299363878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,128,0.003818666769398583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,128,0.007104888558387756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,64,0.0034959999223550162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,64,0.007162666983074612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,16384,128,0.005044444567627377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,16384,32,0.003650666524966558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,16384,32,0.007148444652557373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,65536,0.4021715588039822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,65536,0.2258711126115587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,65536,0.23624444007873535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,16384,0.10300088591045803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,16384,0.05886222256554497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,12288,0.0441964434252845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,12288,0.07789244254430135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,12288,0.0494026674164666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,10240,0.0395084453953637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,10240,0.06685155630111694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,10240,0.04004177782270644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,8192,0.0317191117339664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,8192,0.05412977933883667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,8192,0.033124443557527326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,16384,0.05633955531650119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,7168,0.02863822380701701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,7168,0.04738577869203356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,7168,0.02998222245110406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,6144,0.025418667329682246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,6144,0.04126311010784573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,6144,0.02666666607062022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,5120,0.03510577811135186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,5120,0.023532445232073467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,4096,0.0188711103465822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,4096,0.02942311101489597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,5120,0.022268444299697876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,4096,0.01979200045267741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,3584,0.01715555621518029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,3584,0.025632888078689575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,3584,0.018248889181349013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,3072,0.01571466690964169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,3072,0.02277511192692651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,3072,0.016544888416926067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,2560,0.014020444618331062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,2560,0.01975022256374359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,2560,0.014965333872371249
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,2048,0.016511110795868766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,2048,0.013263999587959714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,1536,0.010933332973056369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,1536,0.013029333617952136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,1536,0.011908444265524546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,1024,0.009167999857001835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,2048,0.012528888881206512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,1024,0.009881777895821465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,1024,0.01036444471942054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,768,0.008384000096056197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,768,0.00833511104186376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,768,0.009721777505344814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,512,0.006787555499209299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,512,0.006620444357395172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,512,0.007370666497283512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,256,0.0052266667286554975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,256,0.004322666674852371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,256,0.006677333265542984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,128,0.0037235555549462638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,128,0.0063386667105886675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,12288,128,0.004869333157936732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,64,0.003379555626047982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,64,0.0063902222447925145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,12288,32,0.003346666693687439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,12288,32,0.006439111298984951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,65536,0.3527368969387478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,65536,0.1849448945787218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,65536,0.2109546661376953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,16384,0.09319200118382771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,16384,0.05369955632421705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,16384,0.04820000131924947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,12288,0.07131910986370511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,12288,0.045174221197764076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,12288,0.03781066669358148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,10240,0.03323644399642944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,10240,0.06342666678958468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,10240,0.0364951127105289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,8192,0.02779200010829502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,8192,0.0524506668249766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,8192,0.030580444468392268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,7168,0.025474665893448725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,7168,0.04785689049296909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,7168,0.02755822241306305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,6144,0.022114665971861944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,6144,0.033738666110568576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,6144,0.024561777710914612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,5120,0.019506666395399306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,5120,0.028528889020284016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,5120,0.02197955548763275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,4096,0.016719111137919955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,4096,0.024086222052574158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,4096,0.018446221947669983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,3584,0.015615110596021017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,3584,0.021381333470344543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,3584,0.017093333933088515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,3072,0.014191110928853353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,3072,0.018924444913864136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,3072,0.015492444237073263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,2560,0.016400888562202454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,1536,0.011031111081441244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,2560,0.014056000444624158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,2048,0.01128888875246048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,2048,0.01370133293999566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,2560,0.01276533305644989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,2048,0.012691555751694573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,1536,0.009993777506881291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,1536,0.011411555939250521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,1024,0.008743111458089616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,1024,0.008548444343937768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,1024,0.010075555907355415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,768,0.008016889293988546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,768,0.007370666497283512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,768,0.009266667068004608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,512,0.006048888795905643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,512,0.006041777630647023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,512,0.007104000283612146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,256,0.005133333305517833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,256,0.0038844446341196695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,256,0.006568888823191325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,10240,128,0.004805333498451444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,128,0.003567999849716822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,128,0.006302222195598815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,64,0.0032195556494924757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,64,0.0063999998900625445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,10240,32,0.0032995556377702286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,10240,32,0.006435555716355641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,16384,0.040475553936428495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,65536,0.2661413351694743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,16384,0.07056711117426555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,16384,0.049676444795396596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,65536,0.1437600056330363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,65536,0.19168444474538168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,12288,0.03125066558519999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,12288,0.055310222837660045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,12288,0.03979111048910353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,10240,0.028864890336990356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,10240,0.047684444321526416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,10240,0.03459466828240289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,8192,0.02471911079353756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,8192,0.03530399998029073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,8192,0.028973334365420874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,7168,0.020983111527231004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,7168,0.03125244379043579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,7168,0.02645066711637709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,6144,0.01866133345497979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,6144,0.02751733362674713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,6144,0.023800000548362732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,5120,0.016700444949997794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,5120,0.023896000451511804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,5120,0.020883555213610332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,4096,0.01443911095460256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,4096,0.020222221811612446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,4096,0.01755466726091173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,3584,0.01352355546421475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,3584,0.01788444485929277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,3584,0.01633777717749278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,3072,0.012431999875439538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,3072,0.015860444969601102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,3072,0.014815110299322339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,2560,0.011319999893506369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,2560,0.01384622189733717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,2560,0.013572444518407186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,2048,0.01016088906261656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,2048,0.011626666618718041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,2048,0.01236088905069563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,1536,0.009545777406957414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,1536,0.010964444114102257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,1024,0.007897777689827813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,1024,0.0074542222751511475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,1024,0.009638222555319468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,768,0.006739555547634761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,768,0.0064604443808396654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,768,0.007631111476156447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,512,0.00553688903649648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,512,0.005427555491526921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,1536,0.009147555463843876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,512,0.0069724445541699724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,256,0.004984000076850255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,256,0.003850666599141227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,256,0.006501333167155583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,8192,128,0.0047155556579430895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,128,0.003458666718668408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,128,0.006216889040337668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,64,0.003245333416594399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,64,0.006288888967699475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,8192,32,0.0033733333564466904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,8192,32,0.006425777657164468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,65536,0.24005778630574545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,65536,0.13280977143181696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,65536,0.13064711623721653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,16384,0.03732622332043118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,16384,0.06435288985570271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,16384,0.03817866577042474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,12288,0.02882400155067444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,12288,0.0446684459845225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,12288,0.03256266646915012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,10240,0.026145777768558923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,10240,0.039043555657068886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,10240,0.026729777455329895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,8192,0.021870222356584337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,8192,0.03228800164328681
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,8192,0.022724444667498272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,7168,0.019900444481107924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,7168,0.028551999065611098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,7168,0.020696888367335003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,6144,0.02495999965402815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,6144,0.018904000520706177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,5120,0.015554666519165039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,5120,0.02176355487770504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,5120,0.017103999853134155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,4096,0.01367288910680347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,4096,0.018426666657129925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,4096,0.014501333236694336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,3584,0.016403555870056152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,3584,0.013639111485746173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,6144,0.01757688820362091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,3072,0.011687111523416309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,3072,0.014552889598740472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,3072,0.012695999609099494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,2560,0.010741333166758219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,2560,0.012539555629094442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,2560,0.011765333513418833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,2048,0.009749333063761393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,2048,0.010839111275143094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,3584,0.012679110798570844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,2048,0.01074044406414032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,1536,0.008843555218643611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,1536,0.009000889129108852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,1536,0.009876444107956356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,1024,0.00759733302725686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,1024,0.007071110937330458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,1024,0.00812533332241906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,768,0.006149333384301927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,768,0.006741333338949416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,512,0.0053484443989064955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,512,0.004554666578769684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,512,0.00629333323902554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,256,0.004879999905824661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,256,0.0035911111368073356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,256,0.0058746664888328975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,768,0.006120000034570694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,7168,128,0.004599110947714912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,128,0.003330666571855545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,128,0.005640000104904175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,64,0.0031315556002987754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,64,0.00572622236278322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,7168,32,0.0032008888406885993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,7168,32,0.00582311095462905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,65536,0.12222845024532741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,65536,0.19656266106499565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,16384,0.033922665648990206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,16384,0.05310222175386217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,65536,0.12424443827735053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,16384,0.0347946650452084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,12288,0.02623822291692098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,12288,0.040956444210476346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,12288,0.02812533246146308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,10240,0.02401422295305464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,10240,0.035262223747041486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,10240,0.024561777710914612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,8192,0.019592000378502738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,8192,0.02889066603448656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,8192,0.02125066684352027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,7168,0.01794488893614875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,7168,0.02576711111598545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,7168,0.01906755566596985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,6144,0.01590133375591702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,6144,0.02258311046494378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,6144,0.01756000022093455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,5120,0.014281777871979607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,5120,0.019845333364274766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,5120,0.015671110815472074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,4096,0.012655999925401477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,4096,0.016844444804721408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,4096,0.013654222091039022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,3584,0.011819555527634092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,3584,0.014940443966123792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,3584,0.012898666991127862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,3072,0.013389333254761167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,3072,0.01203733351495531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,2560,0.011735999749766456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,2560,0.011185777684052786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,3072,0.01089599976936976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,2560,0.010223111344708337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,2048,0.010248888697889116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,2048,0.010233778092596266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,2048,0.009240888886981541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,1536,0.008429333567619324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,1536,0.008584000170230865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,1536,0.009352000223265754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,1024,0.006710222197903528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,1024,0.006649777707126405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,1024,0.007048889166778988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,768,0.005591111050711737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,768,0.0058542220956749385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,768,0.0064737776087390045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,512,0.0052435555391841466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,512,0.0041271112859249115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,512,0.006181333214044571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,256,0.004942222187916438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,256,0.0035706667436493766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,256,0.0059057776298787855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,6144,128,0.0046640001237392426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,128,0.0032951111594835916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,128,0.005603555589914322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,64,0.003087999920050303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,64,0.00564533347884814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,6144,32,0.0031395554542541504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,6144,32,0.005681777579916849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,65536,0.1042906641960144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,65536,0.19163910547892252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,16384,0.030250665214326646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,65536,0.10494933525721233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,16384,0.049025777313444346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,16384,0.030208888981077407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,12288,0.023306666149033442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,12288,0.03840533230039809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,12288,0.02461155586772495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,10240,0.02149066660139296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,10240,0.03367022342152066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,10240,0.02253955602645874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,8192,0.017838221457269456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,8192,0.028188443846172754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,8192,0.019112888309690688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,7168,0.025794666674402025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,7168,0.017092444830470614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,6144,0.019140443868107267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,6144,0.01575644479857551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,7168,0.01644533375898997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,6144,0.014709333578745524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,5120,0.016763556334707472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,5120,0.01442399952146742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,5120,0.01370488852262497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,4096,0.011845333708657159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,4096,0.014466666513019137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,4096,0.012423999607563019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,3584,0.011215111447705163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,3584,0.01253600004646513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,3584,0.01180355581972334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,3072,0.010500444306267632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,3072,0.01112711098459032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,3072,0.011066666907734342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,2560,0.009687110781669617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,2560,0.00999644481473499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,2560,0.010431999961535135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,2048,0.008745777938101027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,2048,0.008671999805503422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,2048,0.009631111390060848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,1536,0.008186666501892937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,1536,0.007479999628331926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,1536,0.008476444416575963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,1024,0.006217777729034424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,1024,0.006112888869312074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,1024,0.006724444528420766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,768,0.005478222337034013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,768,0.0053377776510185665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,768,0.0063164445261160536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,512,0.00509688879052798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,512,0.0038897775941424896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,512,0.006079111248254776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,256,0.004761777818202972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,256,0.0035128887328836652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,256,0.00574311117331187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,5120,128,0.004553777890072929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,128,0.0032346666687064697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,128,0.005516444643338521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,64,0.0030293334275484085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,64,0.005610666755172942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,5120,32,0.003055111194650332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,5120,32,0.005610666755172942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,65536,0.09869422515233357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,65536,0.13896889156765407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,65536,0.08904800150129531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,16384,0.028594666057162817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,16384,0.027215111586782668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,16384,0.038896001047558255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,12288,0.021905778182877436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,12288,0.03114933437771267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,12288,0.021937777598698933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,10240,0.019541333119074505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,10240,0.02700888944996728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,10240,0.019863999552196927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,8192,0.01647111111217075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,8192,0.023904000719388325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,8192,0.017273777061038546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,7168,0.01517599986659156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,7168,0.02197333342499203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,7168,0.01570755574438307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,6144,0.013417777915795645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,6144,0.019713777634832595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,6144,0.014474666780895658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,5120,0.012426666915416718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,5120,0.01696266730626424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,5120,0.013342222405804528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,4096,0.01090133355723487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,4096,0.01481244464715322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,4096,0.01161688897344801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,3584,0.01346577786737018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,3584,0.010282666318946415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,3584,0.011001778145631155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,3072,0.009593778186374241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,3072,0.012459555433856117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,3072,0.010403555300500657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,2560,0.009029332962301042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,2560,0.008753777378135258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,2560,0.009883555273214975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,2048,0.007864000068770515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,2048,0.009120889008045197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,1536,0.006689777804745569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,1536,0.00675999994079272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,2048,0.008287111090289222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,1536,0.0070977773931291364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,1024,0.005768888940413793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,1024,0.005818666683302985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,1024,0.0063528890411059065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,768,0.005391110976537068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,768,0.004190222256713443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,768,0.006123555617200003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,512,0.005095110999213325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,128,0.003080888961752256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,512,0.0037342223028341928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,512,0.005899555567238066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,256,0.004775111046102312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,256,0.0033359999457995095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,256,0.005625777774386936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,4096,128,0.004486222234037188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,128,0.005425777700212266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,64,0.0029173334025674393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,64,0.005492444253630108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,4096,32,0.003137777869900068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,4096,32,0.005532444351249271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,65536,0.1201511091656155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,65536,0.08965510792202419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,65536,0.08183822366926405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,16384,0.026290666725900438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,16384,0.035162667433420815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,16384,0.025892444782786902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,12288,0.020258666740523446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,12288,0.02821777926550971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,12288,0.021179555190934077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,10240,0.01772444446881612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,10240,0.024665777881940205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,10240,0.019095111224386428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,8192,0.021350221501456365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,8192,0.016328000360065036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,7168,0.013936888840463428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,6144,0.014056000444624158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,7168,0.01940444442960951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,8192,0.016029333074887592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,7168,0.015226667126019796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,6144,0.01291911138428582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,6144,0.01739022301303016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,5120,0.011906666888131036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,5120,0.015592889653311836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,5120,0.013054221868515015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,4096,0.013728888498412238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,4096,0.010650667051474253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,4096,0.01127911110719045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,3584,0.010025777750545079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,3584,0.012605333493815528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,3584,0.010840888652536603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,3072,0.009424000150627559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,3072,0.011644444531864591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,3072,0.01020088874631458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,2560,0.008800889054934183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,2560,0.008422222402360704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,2560,0.009696889254781935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,2048,0.007890666524569193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,2048,0.007420444654093848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,2048,0.008404444489214156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,1536,0.006365333166387346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,1536,0.006543111056089401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,1536,0.006878222028414409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,1024,0.005528000079923206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,1024,0.005526222288608551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,1024,0.006281777802440856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,768,0.004098666624890434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,768,0.006114666660626729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,512,0.005040888984998067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,512,0.003656889001528422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,128,0.002991111121243901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,512,0.005879999862776862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,768,0.0053093334039052325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,256,0.0033119999700122406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,256,0.005614222337802251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,128,0.00435644429590967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,128,0.005454222361246745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,64,0.0028373334142896864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,64,0.005502222312821283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3584,32,0.002934222213096089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3584,32,0.005487111293607288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,65536,0.1040231121910943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,65536,0.07759377691480848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,16384,0.02405599918630388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,16384,0.032271110349231295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,16384,0.025194666451878015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3584,256,0.004679999831649992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,65536,0.08011288775338067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,12288,0.01862044466866387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,12288,0.025869333081775244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,12288,0.02032088902261522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,10240,0.016881777180565726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,10240,0.022636445032225713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,10240,0.018372444642914664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,8192,0.014535110857751636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,8192,0.019692444139056735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,8192,0.015790222419632804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,7168,0.013266666895813413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,7168,0.01796888808409373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,7168,0.014656000667148165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,6144,0.012314666476514606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,6144,0.016178665889634024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,6144,0.013613333304723104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,5120,0.01125955581665039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,5120,0.014324444863531323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,5120,0.012672888735930124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,4096,0.01016533292002148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,4096,0.012791111237472959
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,4096,0.01108266661564509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,3584,0.00978133330742518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,3584,0.011952888634469775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,3584,0.010627555350462595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,3072,0.009121778110663096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,3072,0.011109333071443768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,3072,0.009975999593734741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,2560,0.00847999999920527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,2560,0.00794488853878445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,2560,0.00941955546538035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,2048,0.006884444504976273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,2048,0.0071937781241205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,2048,0.007659555309348636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,1536,0.006268444574541516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,1536,0.006346666564544042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,1536,0.006695111178689533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,1024,0.004600888739029567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,1024,0.006294222341643439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,768,0.005222222043408289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,768,0.004064889003833135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,768,0.00612088872326745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,512,0.004954666727119022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,512,0.003690666622585721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,512,0.0058462222417195635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,256,0.0033039999090962936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,256,0.005630222045713001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,128,0.004233777936961916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,1024,0.005494222044944763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,128,0.0029582221888833572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,3072,256,0.004570666700601578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,128,0.005417777846256892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,64,0.0028444443725877335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,64,0.005483555710977978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,3072,32,0.0028986665937635633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,3072,32,0.005479111025730769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,65536,0.07539021968841553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,65536,0.09836355845133464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,16384,0.022966222630606756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,16384,0.027685332629415724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,65536,0.07398755682839288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,16384,0.02409155501259698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,12288,0.018461333380805123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,12288,0.022112889422310725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,12288,0.019647111495335896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,10240,0.016170667277442086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,10240,0.019561777512232464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,10240,0.017747556169827778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,8192,0.01421333352724711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,8192,0.016794666647911072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,8192,0.015249777171346875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,7168,0.013064000341627332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,7168,0.015403555499182807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,7168,0.01424800025092231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,6144,0.012068444656001197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,6144,0.014022222823566861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,6144,0.013362666798962487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,5120,0.011047999891969891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,5120,0.012660444610648684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,5120,0.012430222498046027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,4096,0.009933333430025313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,4096,0.011384000380833944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,4096,0.010792888700962067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,3584,0.009489778015348647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,3584,0.010508444574144153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,3584,0.010363555616802638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,3072,0.00888355573018392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,3072,0.01000088867213991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,3072,0.009713778065310584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,2560,0.007829333345095316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,2560,0.00777511133088006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,2560,0.008245333201355403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,2048,0.00638488887084855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,2048,0.007000889215204451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,2048,0.007100444700982835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,1536,0.006116444451941385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,1536,0.00609688874748018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,1536,0.006647999915811751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,1024,0.005436444448100195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,1024,0.004479111068778568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,1024,0.006217777729034424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,768,0.005129777722888523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,768,0.004007110993067424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,768,0.006033777776691649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,512,0.004882666799757215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,512,0.0036346668170558084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,512,0.0058053334554036455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,256,0.004484444442722532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,256,0.003277333246337043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,256,0.005618666609128316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,128,0.002987555538614591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,128,0.005424889011515512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,64,0.0028151110228565005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,64,0.005375999957323074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2560,32,0.0028728888266616394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2560,32,0.005457777943876054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,65536,0.07389955388175116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,16384,0.023225777679019507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,65536,0.07124622000588311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,16384,0.024336000283559162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,16384,0.023027555810080633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,12288,0.018760000665982563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,12288,0.018909333480728995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2560,128,0.004254222330119875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,10240,0.016511110795868766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,65536,0.07798133293787639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,12288,0.017785777648289997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,10240,0.01716088917520311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,10240,0.01705511079894172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,8192,0.01402311192618476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,8192,0.015030221806632148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,8192,0.014732443624072604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,7168,0.013619555367363824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,7168,0.013815110756291283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,7168,0.01388177772363027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,6144,0.011523555550310345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,6144,0.012808889150619507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,6144,0.012988444831636218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,5120,0.010647999743620554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,5120,0.011707555916574268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,5120,0.01202666676706738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,4096,0.009583111438486312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,4096,0.010514666636784872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,4096,0.010512000156773461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,3584,0.009095999929640029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,3584,0.008940444224410588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,3584,0.009762666291660732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,3072,0.007859555383523306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,3072,0.008192888564533657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,3072,0.008821333448092142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,2560,0.007499555746714274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,2560,0.007609777980380588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,2048,0.006515555497672822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,2560,0.007177777588367462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,2048,0.006812444577614467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,2048,0.006903111106819577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,1536,0.006092444476154115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,1536,0.005683555371231503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,1536,0.006573333508438534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,1024,0.005506666584147348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,1024,0.004376888689067629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,1024,0.006116444451941385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,768,0.005256888767083486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,768,0.004005333201752769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,768,0.005963555640644497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,512,0.004959110998445087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,512,0.003663111064169142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,512,0.00572533326016532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,256,0.004621333132187526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,256,0.003286222202910317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,256,0.005497777627574072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,2048,128,0.004409777621428172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,128,0.0029680000411139596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,128,0.0053839998112784485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,64,0.002809777855873108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,64,0.005356444252861871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,2048,32,0.00286311118139161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,2048,32,0.005404444618357553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,65536,0.07085422012541029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,65536,0.06048266755210029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,65536,0.06910222106509738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,16384,0.01962222158908844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,16384,0.022293332550260756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,16384,0.02130844526820713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,12288,0.017165333032608032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,12288,0.01590844492117564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,12288,0.018339556124475267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,10240,0.015180443723996481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,10240,0.014168888330459595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,10240,0.01660355594423082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,8192,0.013187555803192986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,8192,0.012497777740160624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,8192,0.014320888453059726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,7168,0.012343111137549082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,7168,0.011749332977665795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,7168,0.013528889252079858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,6144,0.011314666933483548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,6144,0.01089777797460556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,6144,0.012714666624863943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,5120,0.010399999717871347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,5120,0.01000355515215132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,5120,0.011783999701340994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,4096,0.00907733374171787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,4096,0.009303111169073317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,4096,0.00941777808798684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,3584,0.007992000215583378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,3584,0.00908622228437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,3584,0.008739555875460306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,3072,0.007236444287829929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,3072,0.008783111141787635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,3072,0.007808888951937358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,2560,0.006711111300521427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,2560,0.007214222517278459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,2560,0.00742133292886946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,2048,0.006243555496136348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,2048,0.006078222145636876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,2048,0.006962666908899943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,1536,0.0058142224119769196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,1536,0.004973333328962326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,1536,0.006568000134494569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,1024,0.0052684446175893145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,512,0.003560888684458203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,1024,0.004257777912749184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,1024,0.006165333506133821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,768,0.005071111023426056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,768,0.003882666842805015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,768,0.005958222266700532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,512,0.0048142220411035745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,512,0.0057297779454125305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,256,0.003176888864901331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,256,0.005517333332035277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,128,0.0029111111329661477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,128,0.005440000030729506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,64,0.002777777819169892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,256,0.004440000073777305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,64,0.005375999957323074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1536,32,0.002850666642189026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1536,32,0.005402666827042897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,65536,0.0709857808219062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1536,128,0.004250666747490565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,65536,0.04231111208597819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,65536,0.06784533129798041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,16384,0.021114667256673176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,16384,0.015437333120240105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,16384,0.02187999917401208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,12288,0.0169493340783649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,12288,0.01310755560795466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,12288,0.01811022228664822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,10240,0.01499466598033905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,10240,0.012133333418104382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,10240,0.016366221838527255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,8192,0.013077333569526672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,8192,0.010766222245163388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,8192,0.014114666316244336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,7168,0.011986666255527072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,7168,0.009898666706350114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,7168,0.013251555462678274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,6144,0.010592000352011787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,6144,0.010579555398888059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,6144,0.011653333074516721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,5120,0.009175111022260454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,5120,0.009891555541091496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,5120,0.010280888941552904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,4096,0.008082666330867344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,4096,0.009184000392754873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,4096,0.008805333740181392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,3584,0.0076248885856734375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,3584,0.008718222379684448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,3584,0.008178667061858708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,3072,0.007151111132568783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,3072,0.008169777691364288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,3072,0.007785778078767989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,2560,0.006612444503439798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,2560,0.0069004446268081665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,2560,0.00739911115831799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,2048,0.006100444330109491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,2048,0.005623999983072281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,2048,0.0069013333155049225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,1536,0.005669333454635408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,1536,0.004983110974232356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,1536,0.006516444600290722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,1024,0.005133333305517833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,1024,0.004284444368547863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,1024,0.006157333238257303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,768,0.004966222163703707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,768,0.003911111089918348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,768,0.005942222144868638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,512,0.0046977777447965415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,512,0.003536000019974179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,512,0.005701333284378052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,256,0.004326222257481681
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,256,0.0031439999325407874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,32,0.002797333316670524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,256,0.005523555394675996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,128,0.0028702221396896574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,128,0.005382222019963794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,1024,64,0.002761777697337998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,64,0.00535377777285046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,1024,128,0.004182222402758068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,1024,32,0.005392888767851724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,65536,0.07029333379533556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,65536,0.03499644332461887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,65536,0.06753955284754436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,16384,0.020742222666740417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,16384,0.01276266657643848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,16384,0.021732444564501446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,12288,0.01682133310370975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,12288,0.010990222295125326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,12288,0.01789955629242791
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,10240,0.014877332581414117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,10240,0.01034577770365609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,10240,0.016138666205936007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,8192,0.012472888661755456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,8192,0.00960355583164427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,8192,0.01365066650840971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,7168,0.010496888723638324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,7168,0.012044444680213928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,6144,0.01000711073478063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,6144,0.009431111315886179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,6144,0.011002666420406766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,5120,0.009011555876996782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,7168,0.01126488877667321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,5120,0.008941333327028487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,5120,0.009723555710580613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,4096,0.008189332981904348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,4096,0.008615111311276754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,3584,0.007684444387753804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,3584,0.008216000265545314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,4096,0.008012444608741337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,3072,0.007087111473083496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,3072,0.0069715554515520734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,3584,0.00758844448460473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,3072,0.007796444826655918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,2560,0.006271111054552927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,2560,0.007400888535711501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,2048,0.006063111126422882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,2048,0.005574222240183089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,2048,0.006931555353932911
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,1536,0.005648889061477449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,1536,0.004889777965015835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,1536,0.006567111031876669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,1024,0.00509688879052798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,1024,0.0041626666982968645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,2560,0.006599999964237213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,1024,0.006144000010357962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,768,0.004845333182149463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,768,0.0037955554823080697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,768,0.005952000204059813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,512,0.0034711110509104202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,512,0.005731555736727185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,256,0.004278222305907143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,256,0.0030951110853089225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,256,0.00554044461912579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,128,0.004127999974621667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,128,0.002871111035346985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,128,0.005393777870469623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,768,512,0.004671999977694618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,64,0.0027057776848475137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,64,0.005387555393907759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,768,32,0.0027866665687825945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,768,32,0.005376889059940974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,65536,0.026187555657492742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,65536,0.07050577799479167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,65536,0.06675200329886542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,16384,0.020786666207843356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,16384,0.010795555180973478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,16384,0.02160888910293579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,12288,0.016531555189026725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,12288,0.010378667049937779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,12288,0.017663111289342243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,10240,0.01427288850148519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,10240,0.009687110781669617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,10240,0.015305777390797934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,8192,0.011848889291286469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,6144,0.009936889012654623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,8192,0.008871999879678091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,8192,0.012065778175989786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,7168,0.011031111081441244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,7168,0.008527999950779809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,7168,0.011769777370823754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,6144,0.007882666256692674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,6144,0.010453333457310995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,5120,0.008936000367005667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,5120,0.0075937774446275495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,5120,0.009678222239017487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,4096,0.008012444608741337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,4096,0.006813333266311222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,4096,0.008551999926567078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,3584,0.007585778004593319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,3584,0.00720000018676122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,3584,0.00815288888083564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,3072,0.007028444773621029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,3072,0.006859555426571105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,3072,0.007797333101431529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,2560,0.00646666685740153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,2560,0.0061919999619325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,2560,0.00737955586777793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,2048,0.006035555568006303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,2048,0.00545866663257281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,2048,0.0069288888739215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,1536,0.005662222289376789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,1536,0.004804444395833545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,1536,0.0064995553758409284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,1024,0.005061333378156026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,1024,0.004086222085687849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,1024,0.006127999888526068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,768,0.004815111143721474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,768,0.003726222034957674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,768,0.005921777751710679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,512,0.004503111044565837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,512,0.0033893332713180115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,512,0.005710222240951326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,256,0.004259555704063839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,256,0.003031111011902491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,256,0.005471999860472149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,512,128,0.004050666673315896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,128,0.0028124445428450904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,128,0.005330666485759947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,64,0.002687999978661537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,64,0.005370666583379109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,512,32,0.0026951111439201566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,512,32,0.005379555539952383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,65536,0.06985866361194186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,65536,0.018399111098713346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,65536,0.06663733058505587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,16384,0.019882667395803664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,16384,0.008976889153321585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,16384,0.020457777712080214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,12288,0.015942222542232938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,12288,0.008219555848174626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,12288,0.016093333562215168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,10240,0.013863110707865821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,10240,0.007517333659860823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,10240,0.013914666242069669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,8192,0.011863110793961419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,8192,0.007206222249401941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,8192,0.011976888610257043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,7168,0.010910222099887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,7168,0.00692622239391009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,7168,0.011215999722480774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,6144,0.009924444059530893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,6144,0.006592888798978593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,6144,0.01037333326207267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,5120,0.008953777452309927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,5120,0.0070426662762959795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,5120,0.009630222287442949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,4096,0.00795022232664956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,4096,0.006680888848172293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,4096,0.00853066643079122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,3584,0.007507555186748505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,3584,0.007113777928882175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,3584,0.00813600007030699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,3072,0.006990222467316522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,3072,0.006772444479995304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,3072,0.007763555480374231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,2560,0.0064604443808396654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,2560,0.006135111053784688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,2560,0.007321777443091075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,2048,0.005930666708283954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,2048,0.005442666510740916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,2048,0.006864000111818314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,1536,0.005600888695981767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,1536,0.004753777964247597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,768,0.0037048889531029593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,1536,0.006485333459244833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,1024,0.005040888984998067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,1024,0.00407644444041782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,1024,0.006050666587220297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,768,0.004776000148720211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,768,0.005869333528810077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,512,0.004513777792453766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,512,0.0033946666452619764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,512,0.005643555687533484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,256,0.00423911131090588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,256,0.002993777808215883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,256,0.005435555759403441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,256,128,0.003998222036494149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,128,0.0028213332924577924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,128,0.005323555734422471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,64,0.0026773332307736077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,64,0.0052826665341854095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,256,32,0.0026497778793176017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,256,32,0.005240889059172736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,65536,0.07041422526041667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,65536,0.014680888917711047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,65536,0.06608266962899102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,16384,0.01963644391960568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,16384,0.007062222394678328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,16384,0.019306666321224637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,12288,0.01590755581855774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,12288,0.00673333348499404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,12288,0.015453333655993143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,10240,0.013807999591032663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,10240,0.0068986668354935115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,10240,0.0138568886452251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,8192,0.011757333245542316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,8192,0.006774222271309958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,8192,0.011983111500740051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,7168,0.010880000061459012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,7168,0.006690666493442323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,7168,0.011178666518794166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,6144,0.009823111196359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,6144,0.006521777974234686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,6144,0.01036444471942054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,5120,0.008907555705971187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,5120,0.0069217777086628815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,5120,0.009579555855857002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,4096,0.007877333296669854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,4096,0.0064560001095136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,4096,0.008485332959228093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,3584,0.007445333732499017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,3584,0.006968888971540663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,3584,0.008117333054542542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,3072,0.0069288888739215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,3072,0.006617777877383762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,3072,0.007698666718271043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,2560,0.006425777657164468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,1536,0.0047271110945277745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,2560,0.006065777606434292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,2560,0.007330666813585494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,2048,0.0058906666106647914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,2048,0.005434666656785541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,2048,0.006825777805513806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,1536,0.0064399999876817065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,1024,0.004047111090686587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,1024,0.006046222315894232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,768,0.004670222186379963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,768,0.003677333394686381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,768,0.0058471109304163195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,1024,0.005007110950019625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,512,0.004459555364317364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,512,0.00332266671790017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,512,0.005670222143332164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,256,0.0041759999261962045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,1536,0.005511111269394557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,256,0.002954666606254048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,256,0.005450666778617435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,48,128,128,0.003933333274390963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,128,0.002805333377586471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,128,0.0052800000541739995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,64,0.0026551110463009942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,64,0.005305777821275923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,128,32,0.002622222320901023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,128,32,0.0052586665583981415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,65536,0.013229333692126803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,65536,0.06657066610124376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,16384,0.006686222222116258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,16384,0.01883555617597368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,12288,0.006542222367392645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,12288,0.015392000476519266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,10240,0.006717333363162146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,10240,0.013826666606797112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,8192,0.006599111275540457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,8192,0.012015110916561551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,7168,0.006524444454246097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,7168,0.011149333582984077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,6144,0.0064284445510970215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,6144,0.01034755590889189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,5120,0.006788444601827198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,5120,0.009611555271678502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,4096,0.0063475556671619415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,4096,0.008519111408127679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,3584,0.00685422205262714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,3584,0.008112000094519721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,3072,0.006534222099516127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,3072,0.007737778127193451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,2560,0.00601244428091579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,2560,0.007320000065697565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,2048,0.0053484443989064955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,2048,0.006912888752089606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,1536,0.0046693334976832075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,1536,0.00646666685740153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,1024,0.003999999827808804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,1024,0.006040888941950268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,768,0.0036071112586392295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,768,0.005859555469618903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,512,0.0033004445334275565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,512,0.00565155554148886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,256,0.0029519999192820657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,256,0.005374222166008419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,128,0.0028080000645584534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,128,0.0052728888889153796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,64,0.002587555597225825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,64,0.0052577778697013855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,64,32,0.0026293332791990707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,64,32,0.0052586665583981415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,65536,0.012527111503813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,65536,0.06605688730875652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,16384,0.006614222294754452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,16384,0.01885066595342424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,12288,0.006409777949253718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,12288,0.015408888459205627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,10240,0.006642666541867786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,10240,0.013824000126785703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,8192,0.006583999842405319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,8192,0.011918221910794577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,7168,0.006488889041874144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,7168,0.011157333023018308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,6144,0.006346666564544042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,6144,0.01037333326207267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,5120,0.0067528887755341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,5120,0.009604444106419882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,4096,0.006313777632183499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,4096,0.008437333007653555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,3584,0.006722666737106111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,3584,0.008057778080304464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,3072,0.00638133328821924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,3072,0.007699555820888943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,2560,0.005997333261701796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,2560,0.007319110963079665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,2048,0.0053048887186580235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,2048,0.006813333266311222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,1536,0.004679111142953237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,1536,0.0064444442590077715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,1024,0.003932444585694207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,1024,0.006039111150635614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,768,0.003552888830502828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,768,0.0058648888435628675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,512,0.003261333331465721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,512,0.005568888866239124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,256,0.0029848888516426086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,256,0.005398222141795688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,128,0.002704888789190186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,128,0.0052684446175893145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,64,0.002550222186578645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,64,0.0053048887186580235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,48,32,32,0.002558222247494592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,48,32,32,0.005225777626037598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,16384,0.5279351340399848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,16384,0.26880354351467556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,16384,0.2953111065758599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,12288,0.21422933207617867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,12288,0.3951831128862169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,12288,0.2307093408372667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,10240,0.3298746744791667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,10240,0.16721244653066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,8192,0.2698728773328993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,8192,0.1453004413180881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,10240,0.18813510735829672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,7168,0.24271821975708008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,8192,0.15829155180189344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,7168,0.1182017723719279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,6144,0.20015645027160645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,7168,0.12866932815975612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,6144,0.11091999875174628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,6144,0.11958844131893581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,5120,0.16773245069715711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,4096,0.13584533002641466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,5120,0.09367822276221381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,5120,0.08741511238945855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,4096,0.07212444146474202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,4096,0.07744977871576945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,3584,0.06596000327004327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,3584,0.11807466877831353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,3584,0.06886044475767347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,2560,0.04795111219088236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,3072,0.05723022090064155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,3072,0.10316266616185506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,3072,0.06257955233256023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,2560,0.08582666847440933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,2560,0.0508266654279497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,2048,0.06977688603931002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,2048,0.04026133484310574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,2048,0.04318399892912971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,1536,0.03185688787036472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,1536,0.05358844333224826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,1536,0.0368115570810106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,1024,0.023678221636348303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,1024,0.03814399904674954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,1024,0.025860443711280823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,768,0.01976355579164293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,768,0.029447111818525527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,768,0.022231111923853557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,512,0.01608088943693373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,512,0.021282666259341772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,512,0.01868355605337355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,256,0.012795555922720166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,256,0.01201600001917945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,256,0.015844444433848064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,65536,128,0.009837333526876237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,128,0.007910221815109253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,128,0.01421955558988783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,64,0.006520888871616787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,64,0.011323555476135679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,65536,32,0.005033777819739448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,65536,32,0.011332444846630096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,65536,0.53566222720676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,65536,0.278876437081231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,16384,0.08448888858159383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,65536,0.33660978741115993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,16384,0.12586577733357748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,16384,0.06990400287840101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,12288,0.05456533365779453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,12288,0.09583822223875259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,12288,0.07031910949283175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,10240,0.04784355560938517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,10240,0.08147466844982572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,10240,0.055989331669277616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,8192,0.06571733289294772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,8192,0.045653333266576133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,7168,0.0342497792508867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,7168,0.059325330787234835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,7168,0.040766222609413996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,8192,0.03875199953715006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,6144,0.050387554698520236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,6144,0.03628888726234436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,5120,0.04275644487804837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,5120,0.026825777358478967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,5120,0.031743112537595965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,4096,0.02273066673013899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,3584,0.020815999971495736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,4096,0.03530844383769565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,4096,0.025570667452282373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,6144,0.03051466743151347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,3584,0.030957334571414526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,3584,0.023984000086784363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,3072,0.018616888258192275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,3072,0.02722399930159251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,3072,0.021592888567182753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,2560,0.016433777080641854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,2560,0.02345955537425147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,2560,0.019204444355434842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,2048,0.014428445034556918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,2048,0.019567110472255282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,2048,0.01741422216097514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,1536,0.012159110771285163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,1536,0.015583111180199517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,1536,0.014163555370436775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,1024,0.010115555591053432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,1024,0.011219555305110084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,1024,0.01200355589389801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,768,0.009105777574910058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,768,0.00923466682434082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,768,0.010976889067225985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,512,0.008000888758235509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,512,0.007325333025720384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,512,0.009903999666372934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,256,0.00543200017677413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,256,0.005109333329730564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,256,0.007349333001507654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,16384,128,0.004956444518433677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,128,0.003649777836269803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,128,0.007066667079925537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,64,0.0033377777371141645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,64,0.007136888802051544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,16384,32,0.003485333174467087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,16384,32,0.007152000235186682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,16384,0.10614133543438381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,16384,0.0621982216835022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,65536,0.3988453282250299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,65536,0.22435998916625977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,65536,0.24846932623121473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,12288,0.08030844397015043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,12288,0.04816177818510267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,10240,0.039213332864973284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,16384,0.059940444098578564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,10240,0.041321777635150485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,8192,0.055027554432551064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,10240,0.0683893362681071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,8192,0.03433510992262099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,12288,0.04612888892491659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,8192,0.034607112407684326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,7168,0.029791110091739233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,7168,0.04836888776885139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,7168,0.03109244505564372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,6144,0.042120890484915845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,6144,0.027523555689387854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,5120,0.02255200015174018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,5120,0.035629332065582275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,5120,0.024467556013001338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,4096,0.019148444135983784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,4096,0.02908088763554891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,4096,0.020469332734743755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,6144,0.025781333446502686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,3584,0.017613333132531907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,3584,0.025916443930731878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,3584,0.018979556030697294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,3072,0.01609422266483307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,3072,0.02260533306333754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,3072,0.017148445049921673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,2560,0.014711999230914645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,2560,0.019620445039537217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,2560,0.015402666396564908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,2048,0.01256622208489312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,2048,0.016354666815863717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,2048,0.01369866645998425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,1536,0.01092266705301073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,1536,0.012979555461141797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,768,0.008075555165608725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,1536,0.012096000214417776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,1024,0.009105777574910058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,1024,0.009611555271678502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,1024,0.010384000009960597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,768,0.008391111261314815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,768,0.009658666948477427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,256,0.006643555644485686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,512,0.006110222389300664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,512,0.006396444307433234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,512,0.007745777567227681
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,256,0.00518222235971027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,256,0.003925333420435588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,12288,128,0.004830222162935468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,128,0.0035297779573334586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,128,0.006350222147173352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,64,0.0032391111469931076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,64,0.0063777777055899305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,12288,32,0.0032871110985676446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,12288,32,0.0064186664919058485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,16384,0.05353777938418918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,65536,0.3812595473395453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,65536,0.2015777826309204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,16384,0.09111199776331584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,65536,0.1851066615846422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,16384,0.05600711372163561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,12288,0.03946844405598111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,12288,0.07217511203553942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,12288,0.04305333230230543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,10240,0.035291555855009295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,10240,0.06452266375223796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,10240,0.03764444589614868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,8192,0.029735998974906072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,8192,0.049914668003718056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,8192,0.0314337776766883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,7168,0.026245334082179602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,7168,0.045612444480260216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,7168,0.02887644370396932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,6144,0.022101332743962605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,6144,0.034028444025251604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,6144,0.025251555773946974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,5120,0.01944266590807173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,5120,0.02915289004643758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,5120,0.022781333989567224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,4096,0.01685066686736213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,4096,0.02392533255947961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,4096,0.018952000472280715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,3584,0.015657777587572735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,3584,0.0214684440029992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,3584,0.017482666505707633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,3072,0.0140337778462304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,3072,0.01885333326127794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,3072,0.01588444411754608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,2560,0.012726222475369772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,2560,0.016026667422718473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,2560,0.014412444498803882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,2048,0.013608000344700284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,2048,0.012763555679056378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,2048,0.01131199962562985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,1536,0.010911111202504901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,1536,0.011560888753996955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,1024,0.008742222355471717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,1024,0.008384000096056197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,1536,0.010032888915803697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,1024,0.010094222095277574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,768,0.00776711106300354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,768,0.007225777539942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,768,0.009132444030708736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,512,0.005884444548024072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,512,0.0058559998869895935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,512,0.007024889190991719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,256,0.0052471111218134565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,256,0.0037475555307335327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,256,0.006626666833957036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,128,0.0033893332713180115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,128,0.006315555423498154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,64,0.003101333354910215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,64,0.006351999938488007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,10240,128,0.004781333522664176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,10240,32,0.003177777760558658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,10240,32,0.0064497776329517365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,65536,0.2616355684068468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,16384,0.04489422175619337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,16384,0.06958311133914523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,16384,0.05189777745140923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,65536,0.15960710578494602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,12288,0.03403466608789232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,65536,0.18730399343702528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,12288,0.05398755603366428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,12288,0.04253866606288486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,10240,0.02884622083769904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,10240,0.04629244407018026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,10240,0.03419288992881775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,8192,0.025254221426116094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,8192,0.03500888744990031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,8192,0.030247999562157526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,7168,0.0225582222143809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,7168,0.031062222189373438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,7168,0.026576000783178542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,6144,0.019545776976479422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,6144,0.027393778165181477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,6144,0.023961777488390606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,5120,0.01734133395883772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,5120,0.02344266739156511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,5120,0.021459556288189355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,4096,0.01982399986849891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,4096,0.017982222967677645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,3584,0.014127110441525778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,3584,0.017815111411942374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,3584,0.016536888149049547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,3072,0.0125946667459276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,2048,0.010312889185216691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,3072,0.015728889240158927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,3072,0.015074667003419666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,2560,0.013612444202105204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,2560,0.013495999905798169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,2048,0.01144088887506061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,2048,0.012292444705963135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,1536,0.009132444030708736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,1536,0.009405333134863112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,1536,0.011002666420406766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,768,0.006322666588756774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,1024,0.007941332956155142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,1024,0.007369777394665613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,1024,0.009759110709031424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,768,0.006221333311663733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,768,0.007738666401969061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,512,0.005427555491526921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,4096,0.014837332897716098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,512,0.005233777893914117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,2560,0.011341333389282227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,512,0.006917333437336816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,256,0.004994666824738185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,256,0.0036888888312710654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,256,0.006575111299753189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,8192,128,0.004624888714816836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,128,0.0033191111352708605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,128,0.006169777777459886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,64,0.003088888815707631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,64,0.006270222365856171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,8192,32,0.0032675556010670135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,8192,32,0.006256888724035687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,65536,0.2313751114739312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,65536,0.14395822419060603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,65536,0.1466791099972195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,16384,0.041527112325032554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,16384,0.06356533368428548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,16384,0.04004888733228048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,12288,0.03207822309599982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,12288,0.04551377892494202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,12288,0.03379377722740173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,10240,0.028664890262815688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,10240,0.03903022077348497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,10240,0.0277164445983039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,8192,0.02345688806639777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,8192,0.0317466656366984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,8192,0.023394667439990576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,7168,0.021136000752449036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,7168,0.028025776147842407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,7168,0.021406221720907424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,6144,0.018702222241295707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,6144,0.024932444095611572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,6144,0.01939644416173299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,5120,0.016688000824716356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,3584,0.013450666434235044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,5120,0.02156266735659705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,5120,0.01738844480779436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,4096,0.014206222361988492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,4096,0.0181048893266254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,4096,0.014576888746685453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,3584,0.016315556234783597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,3584,0.01385244478782018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,3072,0.014426666829321118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,3072,0.012826667063766055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,2560,0.010840888652536603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,1536,0.008709333009190029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,2560,0.012506666282812754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,2560,0.011880888707107969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,2048,0.009815110928482479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,2048,0.010553778045707278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,3072,0.012033777932325998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,2048,0.010835555692513784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,1536,0.008828444613350762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,1536,0.009816889133718278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,1024,0.007091555330488417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,1024,0.006983111302057902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,1024,0.007988444632954067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,768,0.005906666732496685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,768,0.006063111126422882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,768,0.006788444601827198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,512,0.005421333428886201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,512,0.004430222428507275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,512,0.006308444258239534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,256,0.004914666629499859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,256,0.0034551111360390982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,256,0.005903111149867375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,7168,128,0.004557333472702238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,128,0.0032177778581778207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,128,0.00564444437623024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,64,0.002979555477698644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,64,0.005696889013051987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,7168,32,0.003068444422549672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,7168,32,0.005758222192525864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,65536,0.19484355714586046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,16384,0.04340533415476481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,65536,0.1459208859337701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,65536,0.12062577406565349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,16384,0.053598221805360585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,16384,0.03629866573545668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,12288,0.02973066767056783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,12288,0.040778666734695435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,12288,0.030643555853101943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,10240,0.035381333695517644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,10240,0.025403555896547105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,8192,0.0229120006163915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,8192,0.029014223151736792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,8192,0.021299555897712708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,7168,0.020805334051450092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,7168,0.02550044490231408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,10240,0.026872888207435608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,7168,0.020004444652133517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,6144,0.02258844508065118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,6144,0.01833599971400367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,5120,0.01979022224744161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,5120,0.015757333901193406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,4096,0.0166302224000295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,4096,0.013724444641007317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,3584,0.013056889176368713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,3072,0.011759111450778114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,3584,0.014811555544535318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,3584,0.012891555825869242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,3072,0.01330933305952284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,3072,0.012164444559150271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,2560,0.010538666612572141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,2560,0.011639110743999481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,2560,0.011362666885058085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,2048,0.011794666449228922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,2048,0.010053333308961656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,2048,0.010325333310498131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,1536,0.00850933293501536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,1536,0.008432000047630733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,6144,0.018076444665590923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,5120,0.016019556257459853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,4096,0.013705777625242868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,1536,0.00943022221326828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,1024,0.006559999866618051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,1024,0.006612444503439798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,1024,0.0070462218589252895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,768,0.005853333406978183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,768,0.005677333308590784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,768,0.006531555619504716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,512,0.0053297777970631914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,512,0.003903999924659729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,512,0.006264888991912206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,256,0.004961777892377642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,256,0.003440888805521859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,256,0.005865777946180767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,6144,128,0.004545777622196409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,128,0.0031911111954185697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,128,0.005601777798599667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,64,0.0029519999192820657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,64,0.00564533347884814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,6144,32,0.0030382221771611106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,6144,32,0.005648889061477449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,65536,0.12725510862138537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,65536,0.19003555509779188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,65536,0.10119822290208604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,16384,0.04784889022509257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,16384,0.03812711106406318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,16384,0.030504888958401147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,12288,0.02979733215437995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,12288,0.037837333149380155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,12288,0.024673778149816725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,10240,0.02437155610985226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,10240,0.0351475543446011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,10240,0.021943110558721755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,8192,0.021425777011447485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,8192,0.027442667219373915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,8192,0.018927110566033255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,7168,0.019869332512219746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,7168,0.02495644489924113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,7168,0.017180444465743173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,6144,0.016883555385801528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,6144,0.018887110882335238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,6144,0.01574577722284529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,5120,0.014975999792416891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,5120,0.01650399963061015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,5120,0.014512888259357877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,4096,0.013195555243227216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,4096,0.01420177850458357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,4096,0.01257955531279246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,3584,0.01253600004646513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,3584,0.012435555458068848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,3584,0.01180355581972334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,3072,0.011239111423492432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,3072,0.011187555889288584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,3072,0.011146666275130378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,2560,0.010071111222108206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,2560,0.01038577738735411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,2560,0.009743111001120673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,2048,0.009186666872766282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,1024,0.005959110955397288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,2048,0.00868088917599784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,2048,0.009715555442704095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,1536,0.008100444244013892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,1536,0.0074728892909155945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,1536,0.00869511150651508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,1024,0.006625777731339137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,768,0.005620444400442972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,768,0.005143110950787862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,768,0.00629688882165485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,512,0.005256888767083486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,512,0.0037911112109820047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,512,0.006086222413513396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,256,0.004755555755562252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,256,0.0033253334048721525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,256,0.00574311117331187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,128,0.004448889030350579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,128,0.003089777711364958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,128,0.005551110953092575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,64,0.0028977776981062363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,64,0.005573333137565189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,5120,32,0.002954666606254048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,5120,32,0.005648889061477449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,5120,1024,0.006117333140638139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,65536,0.13483822345733643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,16384,0.036828445063696966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,65536,0.12695022424062094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,16384,0.03841777642567953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,65536,0.0862648884455363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,16384,0.027260444230503503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,12288,0.027262222435739305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,12288,0.030586666531032983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,12288,0.022855111294322546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,10240,0.02808177802297804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,10240,0.019685332973798115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,8192,0.020032889313167997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,8192,0.023396443989541795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,8192,0.017068444026841056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,7168,0.018544889158672757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,7168,0.021239999267790053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,7168,0.01554666625128852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,6144,0.016135111451148987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,6144,0.018985778093338013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,6144,0.014375110467274984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,5120,0.014242667290899487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,5120,0.016348444753222995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,5120,0.013292444248994192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,4096,0.012429333395428128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,4096,0.014211555322011312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,10240,0.02256711158487532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,4096,0.011537777880827585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,3584,0.013027555412716336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,3584,0.01096088853147295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,3584,0.011892444557613797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,3072,0.011912888950771756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,3072,0.010423111418883005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,2560,0.009715555442704095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,2560,0.008736888567606607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,2560,0.009846222069528367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,2048,0.007824000385072496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,3072,0.010696888797812991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,2048,0.008958222137557136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,1536,0.007147555549939473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,1536,0.006728888799746831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,1536,0.007679999702506595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,1024,0.005730666634109285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,1024,0.006315555423498154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,768,0.005511999958091312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,768,0.004097777936193678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,2048,0.008731555607583787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,768,0.006117333140638139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,512,0.0052177777720822235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,512,0.0036755556033717263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,512,0.00589155571328269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,256,0.0047128887640105355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,256,0.0032995556377702286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,256,0.005596444424655702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,128,0.004449777719047335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,128,0.00300177786913183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,128,0.005467555589146084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,64,0.0028613333900769553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,64,0.005474666754404704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,4096,32,0.003060444361633725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,4096,32,0.005516444643338521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,65536,0.12033422787984212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,65536,0.11921955479515924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,16384,0.035383111900753446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,16384,0.034852445125579834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,65536,0.08153333266576131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,4096,1024,0.005988444305128521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,16384,0.02569599946339925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,12288,0.026603556341595117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,12288,0.027688889039887324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,12288,0.020964443683624268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,10240,0.022854222191704646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,10240,0.024289778537220422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,10240,0.0189057770702574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,8192,0.020005333754751418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,8192,0.020793777373101976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,8192,0.0163111107216941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,7168,0.018332444959216647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,7168,0.018840889135996502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,7168,0.015007999208238391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,6144,0.016169778174824186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,6144,0.01681066718366411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,6144,0.013956444131003486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,5120,0.014090667168299357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,5120,0.015033778217103748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,5120,0.012895999683274163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,4096,0.012454222473833295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,4096,0.013233777549531726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,4096,0.011240888800885944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,3584,0.011535999675591787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,3584,0.012201777762836881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,3584,0.010728889041476779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,3072,0.010599111517270407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,3072,0.011192888849311404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,3072,0.010201777848932479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,2560,0.009374221993817223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,2560,0.008263111114501953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,2560,0.009678222239017487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,2048,0.008307555483447181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,2048,0.007357333269384172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,2048,0.008224888808197444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,1536,0.006538666784763336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,1536,0.006521777974234686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,1536,0.007032000356250339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,1024,0.005722666780153911
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,1024,0.00535111129283905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,1024,0.006340444501903322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,768,0.003995555556482739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,768,0.006178666734033161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,512,0.00517866677708096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,512,0.003612444632583194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,512,0.0058782220714622075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,256,0.0046844445168972015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,256,0.0032551110618644287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,256,0.005635555419656966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,32,0.0028968888024489083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,128,0.004430222428507275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,128,0.0029440000653266907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,128,0.00544088871942626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3584,64,0.0027751111321979095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,64,0.005450666778617435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3584,768,0.005388444496525659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3584,32,0.005446222093370225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,65536,0.10281155506769817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,65536,0.10997066895167033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,16384,0.03281777766015794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,16384,0.031679110394583807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,65536,0.07707822322845459
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,16384,0.024680000212457445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,12288,0.025240888198216755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,12288,0.02528000043498145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,12288,0.02011644509103563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,10240,0.021887110339270696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,10240,0.022427555587556627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,10240,0.018201778332392376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,8192,0.01903022163444095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,8192,0.019274666905403137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,8192,0.01568000018596649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,7168,0.017472889688279893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,7168,0.014583999911944071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,6144,0.014528888795110913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,6144,0.015591111448076038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,6144,0.013562666873137156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,7168,0.017010667257838778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,5120,0.013548444542619916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,5120,0.0140364451540841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,5120,0.012697777814335294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,4096,0.01185066666867998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,4096,0.012643555800120035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,4096,0.010958222051461538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,3584,0.011176000038782755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,3584,0.011603555745548673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,3584,0.01054222219520145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,3072,0.010024000373151567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,3072,0.010814222196737925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,3072,0.00997155573632982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,2560,0.007982222570313348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,2560,0.008996444443861643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,2048,0.007456888755162557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,2048,0.007118222614129384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,2048,0.007483555210961237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,2560,0.009204444785912832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,1536,0.00657066661450598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,1536,0.006243555496136348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,1536,0.006820444431569841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,1024,0.005692444327804778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,1024,0.0046640001237392426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,1024,0.006320888797442119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,768,0.005292444593376583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,768,0.00397511116332478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,768,0.006141333530346553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,512,0.005011555635266834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,512,0.0036142224238978494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,512,0.00590844452381134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,256,0.004505777938498391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,256,0.003237333355678452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,256,0.005609777652555042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,128,0.0029235554652081597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,128,0.00544888898730278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,3072,128,0.004222222086456087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,64,0.002785777880085839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,64,0.005490666462315454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,3072,32,0.002874666617976295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,3072,32,0.005489777773618698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,65536,0.09988088740242852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,16384,0.027564444475703772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,65536,0.07320800092485216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,16384,0.028425776296191748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,16384,0.023834667272037927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,12288,0.022350221872329712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,12288,0.02164177762137519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,12288,0.019330667124854196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,10240,0.0188719994492001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,10240,0.019095111224386428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,10240,0.01754044493039449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,8192,0.018033777674039204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,8192,0.016541333662139047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,8192,0.015013333823945789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,7168,0.017070222232076857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,7168,0.015020444989204407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,7168,0.014244443840450711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,6144,0.013927999469969006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,6144,0.013669333524174161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,6144,0.013110222087966071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,5120,0.012528888881206512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,5120,0.012278222375445895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,5120,0.012230222423871359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,4096,0.011236444115638733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,4096,0.01110666659143236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,65536,0.09218488799201117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,4096,0.010760889285140567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,3584,0.010300444232092964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,3584,0.010485333700974783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,3584,0.010223999619483948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,3072,0.0090515555606948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,3072,0.009798222117953831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,3072,0.009672000176376766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,2560,0.008018666671382057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,2560,0.007608888877762689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,2560,0.008409777449236976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,2048,0.0069395556218094295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,2048,0.007188444336255391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,1536,0.006287999865081575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,1536,0.005910222315125995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,1536,0.006643555644485686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,1024,0.00546577779783143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,1024,0.004541333350870344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,1024,0.006234666539563074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,768,0.0052471111218134565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,768,0.003955555458863576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,512,0.004864888886610667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,768,0.006052444378534953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,512,0.003581333491537306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,512,0.0058764442801475525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,2048,0.007072889142566257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,256,0.004481777962711123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,256,0.003216888962520493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,256,0.005538666827811136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,128,0.0028826666788922418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,128,0.005399999933110342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,64,0.0027804445061418745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,64,0.005415110952324337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2560,32,0.0028560000161329904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2560,32,0.005418666534953647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,65536,0.07204177644517687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,65536,0.0706604454252455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,16384,0.030921777089436848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2560,128,0.004113777644104428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,16384,0.023835556374655828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,16384,0.02276800076166789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,12288,0.024903110331959192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,12288,0.019290667441156175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,12288,0.018704000446531508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,10240,0.021372444099850122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,10240,0.017066667477289837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,10240,0.016898666818936665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,8192,0.018421333697107103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,65536,0.10080177916420831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,8192,0.014500444134076437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,8192,0.014663111832406787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,7168,0.016606221596399944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,7168,0.013562666873137156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,7168,0.01368977791733212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,6144,0.012501333322789935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,6144,0.012826667063766055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,5120,0.01314133322901196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,5120,0.011351111034552256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,6144,0.014117333624098035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,5120,0.011845333708657159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,4096,0.011523555550310345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,4096,0.01013777818944719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,4096,0.010271111296282874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,3584,0.01055111073785358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,3584,0.009376889301670922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,3584,0.00979555563794242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,3072,0.008934222161769867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,3072,0.008906666603353288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,3072,0.00848355558183458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,2560,0.008028444316652086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,2560,0.00738933351304796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,2560,0.007541333635648091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,2048,0.007231111327807109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,2048,0.006633777585294511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,2048,0.0069164443347189165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,1536,0.006520000182920032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,1536,0.005161777966552311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,1536,0.006523555351628198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,1024,0.005631111148330901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,1024,0.0041759999261962045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,1024,0.006090666684839461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,768,0.005252444495757421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,768,0.0038488888078265717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,768,0.005929777605666055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,512,0.004967111266321606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,512,0.0034248888906505373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,512,0.005719111197524601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,256,0.004646222210592694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,256,0.003112888791494899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,256,0.005495999836259418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,2048,128,0.004341333276695675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,128,0.0028604444944196274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,128,0.005294222384691238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,64,0.0027671110712819626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,64,0.00535111129283905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,16384,0.027425777581002977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,2048,32,0.002963555562827322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,2048,32,0.005336888962321811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,65536,0.0591742197672526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,65536,0.091103110048506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,65536,0.06879733006159465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,16384,0.01905333333545261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,12288,0.022272888157102797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,16384,0.022110222114457026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,12288,0.015421332584487068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,12288,0.01816177699300978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,10240,0.014008000493049622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,10240,0.016556445095274184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,8192,0.012317332956526013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,8192,0.014357333381970724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,7168,0.014745778507656522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,7168,0.011332444846630096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,7168,0.013501333693663279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,6144,0.01387733303838306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,6144,0.010641777680979835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,6144,0.012558221817016602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,5120,0.009815110928482479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,5120,0.011689778003427716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,4096,0.009761778016885122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,4096,0.008993777963850234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,4096,0.009128889275921715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,10240,0.019703111714786954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,3584,0.009220444493823582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,3584,0.008939555949634975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,8192,0.01706755492422316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,3584,0.008818666968080733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,2560,0.0074240002367231585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,3072,0.008496888809733922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,3072,0.008562666674455007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,5120,0.011718221836619906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,3072,0.007812444534566667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,2560,0.007338666253619724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,2560,0.007128000259399414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,2048,0.006666666517655055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,2048,0.006053333481152852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,2048,0.0069573331210348345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,1536,0.006242666807439592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,768,0.003809777812825309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,1536,0.00489155575633049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,1536,0.006520000182920032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,1024,0.0041991112132867174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,1024,0.006151999864313338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,768,0.005118222286303838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,768,0.005991111199061076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,512,0.0047733332547876565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,512,0.0034755555291970572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,512,0.00573244442542394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,256,0.004397333496146732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,1024,0.005383111122581694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,256,0.0030782222747802734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,256,0.005519111123349931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1536,128,0.0041253334946102565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,32,0.005401777724424998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,128,0.0028817777832349143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,128,0.005387555393907759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1536,64,0.005390222287840313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,64,0.002696000039577484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1536,32,0.002785777880085839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,65536,0.040904889504114784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,16384,0.02082577844460805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,65536,0.06753510899013944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,16384,0.014877332581414117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,10240,0.011801777614487542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,16384,0.02167911165290409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,12288,0.018600889378123812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,12288,0.012768000364303589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,12288,0.01789777808719211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,10240,0.01534400052494473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,10240,0.016179554992251925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,8192,0.013256000147925483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,8192,0.010407999985747868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,8192,0.013975999421543546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,7168,0.01201333353916804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,7168,0.009693333672152625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,7168,0.012834666503800286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,6144,0.010613333019945355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,6144,0.008772444393899705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,6144,0.011390222443474663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,65536,0.07218311230341594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,5120,0.009127111070685917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,5120,0.008755555583371056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,5120,0.010114666488435533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,4096,0.007963555554548899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,4096,0.008210666477680206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,4096,0.008567999634477828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,3584,0.007640889121426477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,3584,0.007852444218264686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,3584,0.008181333541870117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,3072,0.007110222346252865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,3072,0.007277333074145847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,3072,0.007805333369308048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,2560,0.006510222123728857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,2560,0.006440889090299606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,2560,0.007370666497283512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,2048,0.006175999840100606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,2048,0.005502222312821283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,2048,0.0068915556702348925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,1536,0.005647999958859549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,1536,0.004776000148720211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,1536,0.006495111104514863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,1024,0.005139555368158552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,1024,0.004108444270160464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,1024,0.00610133343272739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,768,0.004982222285535601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,768,0.003714666598372989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,768,0.005943999936183293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,512,0.0046568889584806235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,512,0.0034142221427626083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,512,0.005721777677536011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,256,0.004281777888536453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,256,0.0030417777597904205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,256,0.005473777651786804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,1024,128,0.004256000121434529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,128,0.0028444443725877335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,128,0.0053635554181204895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,64,0.0027022223091787765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,64,0.0053226666318045715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,1024,32,0.002736888825893402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,1024,32,0.00535111129283905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,65536,0.03371555606524149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,65536,0.06710755825042725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,16384,0.01236088905069563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,16384,0.02154133386082119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,12288,0.016691555579503376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,12288,0.010582221878899468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,12288,0.01772711177666982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,10240,0.014775110615624322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,10240,0.010103999740547603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,10240,0.01605866683853997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,8192,0.012332444389661154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,8192,0.009372444616423713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,8192,0.01314133322901196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,7168,0.008776889079146916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,7168,0.011678222152921887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,6144,0.008295111358165741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,6144,0.010667555862002902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,5120,0.008911111288600499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,65536,0.06966311401791044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,5120,0.008416888614495596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,5120,0.009693333672152625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,4096,0.007239110767841339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,4096,0.008567999634477828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,16384,0.020463110672103036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,7168,0.01089599976936976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,6144,0.009911110831631554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,3584,0.007522666619883643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,3584,0.007536888950400882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,3584,0.008192000289758047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,4096,0.007995555798212687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,3072,0.006974221931563483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,3072,0.0069546666410234236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,3072,0.007781333393520779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,2560,0.006544000158707301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,2560,0.006155555446942647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,2560,0.007418666448858049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,2048,0.00608711110221015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,2048,0.005439110928111606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,2048,0.006893333461549547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,1536,0.0047422221137417685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,1536,0.006541333264774746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,1024,0.005032000028424793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,1024,0.004063999901215236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,512,0.0034222222036785553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,1024,0.006157333238257303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,768,0.0036968886852264404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,768,0.00590844452381134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,512,0.0046364445653226645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,1536,0.005561777700980504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,512,0.005738666488064661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,256,0.004300444490379757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,256,0.003031999907559819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,768,0.004844444493452708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,256,0.005481777919663324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,768,128,0.004115555435419083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,128,0.0028435554769304064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,128,0.0053671110007498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,64,0.002704000100493431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,64,0.005373333477311664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,768,32,0.0026951111439201566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,768,32,0.005351999981535806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,65536,0.06974399752087064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,65536,0.02516799999607934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,65536,0.06656266583336724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,16384,0.02033155494266086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,16384,0.010427555276287926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,16384,0.02147377696302202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,12288,0.016354666815863717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,12288,0.009203555683294931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,12288,0.017210667332013447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,10240,0.013931555880440606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,10240,0.009007111191749573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,10240,0.014831110835075378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,8192,0.011818666425016193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,8192,0.008373333348168267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,8192,0.0120319997270902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,7168,0.010834666589895884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,7168,0.008084444536103142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,7168,0.01132266637351778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,6144,0.007341333561473423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,6144,0.010443555812040964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,5120,0.008866666919655269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,5120,0.007300444775157505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,5120,0.009647111097971598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,4096,0.00793599999613232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,4096,0.006883555402358373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,4096,0.00850133349498113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,3584,0.0075048887067370946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,3584,0.007103111180994246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,3584,0.008135110967689091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,3072,0.007024889190991719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,3072,0.006725333217117522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,6144,0.009865778187910715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,3072,0.007740444607204861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,2560,0.006504888749784893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,2560,0.006121777825885349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,2560,0.007376000285148621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,2048,0.006009777800904379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,2048,0.005429333282841577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,2048,0.006882666713661618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,1536,0.005601777798599667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,1536,0.00479644454187817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,1536,0.006504888749784893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,1024,0.005012444323963589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,1024,0.00406577769252989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,1024,0.006120000034570694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,768,0.004772444566090901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,768,0.003702222059170405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,768,0.005931555396980709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,512,0.004489777816666497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,512,0.0035048888789282907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,512,0.005670222143332164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,256,0.004256000121434529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,256,0.003012444410059187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,256,0.005423999908897612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,512,128,0.0041493334703975255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,128,0.002850666642189026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,128,0.0053288886944452924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,64,0.0026968889352348116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,64,0.005344888816277186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,512,32,0.0026844443960322272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,512,32,0.0053173332578606075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,65536,0.0694355567296346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,65536,0.017645332548353408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,65536,0.06639377938376533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,16384,0.01958488921324412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,16384,0.00847111145655314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,16384,0.02017688916789161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,12288,0.015704888436529372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,12288,0.007376888559924231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,12288,0.015798222687509324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,10240,0.013703111145231458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,10240,0.007430222299363878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,10240,0.01384888920519087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,8192,0.011698666546079846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,8192,0.0068488890926043195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,8192,0.01202133297920227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,7168,0.01072088877360026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,7168,0.006734222173690796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,7168,0.011215999722480774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,4096,0.007892444729804993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,6144,0.0064693333374129395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,6144,0.010332444475756751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,5120,0.008857777549160851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,5120,0.0069253332912921906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,5120,0.009580444130632613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,4096,0.006539555473460092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,4096,0.008466666771305932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,3584,0.007452444069915348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,3584,0.006884444504976273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,3584,0.008120888637171851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,3072,0.006912888752089606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,3072,0.006582222051090664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,3072,0.007713778151406183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,2560,0.006451555424266391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,2560,0.006075555665625467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,2560,0.007416889071464539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,2048,0.005973333285914526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,2048,0.005416000054942236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,2048,0.006836444553401735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,1536,0.005519111123349931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,1536,0.0047288888858424295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,1536,0.006423111177153057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,1024,0.004986666556861666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,6144,0.009771555662155151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,1024,0.00407644444041782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,1024,0.006086222413513396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,768,0.004782222211360931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,768,0.003677333394686381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,768,0.005858666780922148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,512,0.004426666845877965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,512,0.003350222276316749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,512,0.005619555711746216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,256,0.004146666576464971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,256,0.0029893333299292457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,256,0.005410666680998272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,256,128,0.00397511116332478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,128,0.0028231110837724474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,128,0.00527022240890397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,64,0.0026666666898462507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,64,0.005277333160241445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,256,32,0.002672000063790215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,256,32,0.005256888767083486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,65536,0.06898399856355455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,65536,0.014489778214030795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,65536,0.06612355841530694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,16384,0.01938666734430525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,16384,0.006847111301289664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,16384,0.01889866590499878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,12288,0.015641777051819693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,12288,0.006578666468461354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,12288,0.01535377734237247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,10240,0.013653332988421122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,10240,0.006678222368160884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,10240,0.013791999883121915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,8192,0.011735111474990845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,8192,0.006611555400821898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,8192,0.011959111524952782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,7168,0.01071200023094813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,7168,0.006519111080302133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,7168,0.011156444748242697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,6144,0.009782222410043081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,6144,0.0064444442590077715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,6144,0.010416000253624385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,5120,0.008763555851247575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,5120,0.006814222368929122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,5120,0.009740444521109263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,4096,0.007816000117195977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,4096,0.006361777583758037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,4096,0.008466666771305932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,3584,0.007370666497283512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,3584,0.00684533350997501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,3584,0.008064000142945183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,3072,0.0069288888739215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,3072,0.006535111202134027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,3072,0.007679999702506595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,2560,0.0063742221229606206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,2560,0.006031111296680238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,2560,0.007276444799370236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,2048,0.005924444645643234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,2048,0.0053377776510185665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,2048,0.006842666616042455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,1536,0.005440000030729506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,1536,0.004650666481918759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,1536,0.0064177778032090925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,1024,0.004972444640265571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,1024,0.00398577791121271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,1024,0.006034666465388403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,768,0.0036231109665499795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,768,0.005853333406978183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,512,0.0032755554550223877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,512,0.005586666779385672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,256,0.004132444245947732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,256,0.0029315555261241067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,256,0.005415110952324337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,128,0.003986666599909465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,128,0.0027635554886526535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,128,0.0052426668504873914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,64,0.002587555597225825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,64,0.0052595556610160405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,65536,0.06619822316699557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,128,32,0.002611555573013094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,128,32,0.005241777747869492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,65536,0.012865777644846173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,16384,0.006550222221348021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,16384,0.018888889087571036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,12288,0.006414222220579783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,12288,0.015416888727082146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,10240,0.006601777755551868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,10240,0.013788444300492605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,8192,0.006492444624503453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,8192,0.011961778004964193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,768,0.0047324444684717385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,32,128,512,0.004407111141416761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,7168,0.006365333166387346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,7168,0.011149333582984077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,6144,0.006193777753247156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,6144,0.010374222364690568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,5120,0.006520000182920032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,5120,0.009583111438486312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,4096,0.006128888991143968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,4096,0.008468444148699442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,3584,0.006602666858169768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,3584,0.008091555701361762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,3072,0.006319111006127463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,3072,0.007697777615653143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,2560,0.005963555640644497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,2560,0.007310222420427535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,2048,0.005315555466545953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,2048,0.006798222247097228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,1536,0.004646222210592694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,1536,0.0064604443808396654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,1024,0.003998222036494149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,1024,0.006041777630647023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,768,0.003563555578390757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,768,0.0058604445722368025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,512,0.00328533330725299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,512,0.005629333357016246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,256,0.0028880000528362063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,256,0.005371555685997009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,128,0.0027751111321979095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,128,0.005254222287072076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,64,0.0025395556456512874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,64,0.0052577778697013855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,64,32,0.002558222247494592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,64,32,0.0052328887912962176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,65536,0.012582222620646158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,65536,0.06537599696053399
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,16384,0.006342222293217977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,16384,0.018834667073355783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,12288,0.006666666517655055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,12288,0.015384000208642749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,10240,0.006516444600290722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,10240,0.013727111121018728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,8192,0.0063893331421746155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,8192,0.011956444217099084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,5120,0.009503111243247986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,7168,0.006181333214044571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,7168,0.011113777756690979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,6144,0.006098666538794835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,6144,0.010330666270520952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,5120,0.006439111298984951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,4096,0.006066666709052191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,4096,0.00849422232972251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,3584,0.006510222123728857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,3584,0.008085333638721043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,3072,0.006116444451941385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,3072,0.007720888488822513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,2560,0.005958222266700532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,2560,0.007311999797821045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,2048,0.0052888890107472735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,2048,0.006853333363930385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,1536,0.004581333448489507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,1536,0.0063786668082078295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,1024,0.003935111065705617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,1024,0.006007110906971826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,768,0.003593777616818746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,768,0.0058044443527857465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,512,0.003269333392381668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,512,0.005583111196756363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,256,0.0029004443850782183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,256,0.0053226666318045715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,128,0.0026684444811609057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,128,0.005209777918126848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,64,0.0025840000145965153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,64,0.005199111170238919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,32,32,32,0.002516444358560774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,32,32,32,0.005207111024194294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,16384,0.5246817800733778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,16384,0.26820354991488987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,12288,0.392070213953654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,12288,0.21323111322191027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,12288,0.2347946696811252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,16384,0.2983022265964084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,10240,0.3285120063357883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,8192,0.2633235454559326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,8192,0.13488088713751897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,7168,0.23071556621127656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,10240,0.1981359985139635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,8192,0.14239911238352457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,7168,0.11840444140964085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,6144,0.19831732908884683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,7168,0.14254132906595865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,5120,0.1669608884387546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,5120,0.08628444539176093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,6144,0.12337066067589654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,4096,0.1340622239642673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,5120,0.10432622167799209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,4096,0.07347466548283894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,4096,0.07690133651097615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,3584,0.0625422199567159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,3584,0.0678960018687778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,3584,0.11731021934085423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,3072,0.10187555683983697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,3072,0.06074133184221056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,2560,0.048080000612470836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,10240,0.16642222139570448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,2560,0.08535466591517131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,2560,0.05087644524044461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,2048,0.039311998420291476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,2048,0.0689200030432807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,2048,0.04237688912285698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,1536,0.034401777717802264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,1536,0.033244444264305964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,1536,0.05267555514971415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,1024,0.03684355484114753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,6144,0.10341511170069377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,1024,0.025898666845427618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,768,0.029177778297000464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,768,0.022318222456508215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,512,0.015632889337009855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,512,0.022327999273935955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,512,0.019374221563339233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,256,0.011976888610257043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,256,0.011500444677140979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,256,0.015857777661747403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,3072,0.05631644195980496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,128,0.009356444080670675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,1024,0.023443554838498432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,128,0.007583111524581909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,65536,768,0.019928889142142404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,128,0.014017777310477363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,64,0.005776888794369168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,64,0.011155555645624796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,65536,32,0.004640000147951974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,65536,32,0.011214222345087262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,65536,0.5301235516866049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,65536,0.28311999638875324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,65536,0.34063466389973956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,16384,0.1241039964887831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,16384,0.07303822040557861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,16384,0.08525688780678643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,12288,0.05487022134992811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,12288,0.09460533327526516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,12288,0.07206577724880643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,10240,0.07961599694357978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,10240,0.050082667006386646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,10240,0.05619821945826212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,8192,0.06458933485878839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,8192,0.045924444993337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,7168,0.056890666484832764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,7168,0.036878221564822726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,7168,0.04154933492342631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,6144,0.04942399925655789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,6144,0.036297778288523354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,5120,0.026571555270089045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,5120,0.04208888941340976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,5120,0.03199999862247043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,4096,0.022582223018010456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,4096,0.034834666384591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,4096,0.026712000370025635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,8192,0.03905866543451945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,3584,0.03054844339688619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,3584,0.02461599972512987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,3072,0.018442667192882962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,6144,0.03039466672473484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,3072,0.026837334036827087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,3072,0.0218551109234492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,2560,0.023090667194790308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,2560,0.019313777486483257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,2048,0.014259555273585849
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,3584,0.020530665914217632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,2048,0.019302222463819716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,2048,0.016584000653690763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,1536,0.012240000069141388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,1536,0.015375999940766228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,1536,0.014371555712487964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,1024,0.010213333699438307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,2560,0.01647911138004727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,1024,0.011054221954610614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,1024,0.011951111257076263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,768,0.009053332938088311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,768,0.009148444566461775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,768,0.011048888994587792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,512,0.008103999826643202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,512,0.007182222273614671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,512,0.00977955593003167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,256,0.00534400012758043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,256,0.00517511119445165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,256,0.007297777467303806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,16384,128,0.005012444323963589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,128,0.0035537779331207275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,128,0.006976000136799282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,64,0.0032257777121331957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,64,0.0069475554757648045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,16384,32,0.0033297776761982176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,16384,32,0.0069982219073507525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,16384,0.1049155526691013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,16384,0.06825244426727295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,16384,0.06210933128992716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,65536,0.42730400297376847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,12288,0.051875554853015475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,12288,0.07855288849936591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,65536,0.2551857895321316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,65536,0.24577956729465059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,12288,0.047651555803087026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,10240,0.0457217792669932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,10240,0.06708088848325941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,10240,0.04532355401251051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,8192,0.03713511096106635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,8192,0.05335022343529595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,8192,0.036481777826944985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,7168,0.032836443848080106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,7168,0.04764622118737963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,7168,0.03127466638882955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,6144,0.028913779391182795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,6144,0.043064888980653554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,6144,0.027497778336207073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,5120,0.02494844463136461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,5120,0.03501510951254103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,5120,0.02454311152299245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,4096,0.02232444451914893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,4096,0.028301331732008193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,4096,0.020626667473051283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,3584,0.018931556079122756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,3584,0.025382222400771245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,3584,0.01887999971707662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,3072,0.022315555148654517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,3072,0.017162667380438913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,2560,0.014932443698247274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,2560,0.019176888797018263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,2560,0.01536799967288971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,2048,0.013027555412716336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,2048,0.01607822212908003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,2048,0.01349422252840466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,1536,0.01108444482088089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,1536,0.012867555850081973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,1536,0.012074666718641916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,1024,0.009165333376990424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,1024,0.009501333038012186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,3072,0.017747556169827778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,1024,0.01052266690466139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,768,0.008310221963458592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,768,0.007879999776681265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,768,0.009645333720578088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,512,0.006042666733264923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,512,0.006326222171386083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,512,0.007862221863534715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,256,0.0052666668262746595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,256,0.0038764443662431505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,256,0.006586666736337874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,12288,128,0.004765333400832282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,128,0.0034560000316964257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,128,0.0063484443558586975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,64,0.003112888791494899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,64,0.006303999986913469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,12288,32,0.003120888852410846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,12288,32,0.006331555545330048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,16384,0.0565297802289327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,65536,0.3708791202969021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,65536,0.20469244321187338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,16384,0.09214578072230022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,16384,0.056200000974867076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,12288,0.04190577732192146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,12288,0.07078578074773152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,12288,0.04343377881579929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,10240,0.03812888926929898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,10240,0.06131466892030504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,10240,0.03775733378198411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,8192,0.03015022145377265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,8192,0.046797331836488515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,65536,0.19531910949283174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,8192,0.030997332599427965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,7168,0.027412444353103638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,7168,0.0440684457619985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,7168,0.02854666776127285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,6144,0.02255200015174018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,6144,0.03940444522433811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,6144,0.02532177832391527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,5120,0.02867822183503045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,5120,0.02090577781200409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,5120,0.02281155520015293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,4096,0.017151110702090792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,4096,0.02370311154259576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,4096,0.018781332506073844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,3584,0.020429333051045735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,3584,0.01614488826857673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,3584,0.017472000585661996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,3072,0.01421866648726993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,3072,0.018632888793945312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,3072,0.015959110524919298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,2560,0.012816888590653738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,2560,0.016120000018013846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,2560,0.014568888478808932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,2048,0.011687111523416309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,2048,0.013276444541083442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,2048,0.012944888737466602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,1536,0.010130667024188572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,1536,0.011720000041855706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,1536,0.011552888486120435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,1024,0.008159110943476358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,1024,0.010219555762079027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,768,0.007882666256692674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,768,0.007116444408893585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,768,0.009296889106432596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,512,0.005924444645643234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,512,0.0057857777509424425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,512,0.007117333511511485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,256,0.003617777592606015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,256,0.005220444252093633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,256,0.006615110983451207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,128,0.004851555658711327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,128,0.003261333331465721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,128,0.006245333287451003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,64,0.003019555575317807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,10240,1024,0.008714666797055138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,64,0.006310222049554189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,10240,32,0.003014222201373842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,10240,32,0.006335111127959357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,65536,0.2579164505004883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,65536,0.17794577280680338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,16384,0.05004355642530653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,65536,0.19582045078277588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,16384,0.07573778099483915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,16384,0.05219911204444038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,12288,0.03769066598680284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,12288,0.05847733550601535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,12288,0.04458489020665487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,10240,0.03090755475891961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,10240,0.049710220760769315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,10240,0.03534400131967332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,8192,0.02646933330429925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,8192,0.03447466757562425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,8192,0.028945777151319716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,7168,0.023396443989541795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,7168,0.030582222673628066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,7168,0.02711822258101569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,6144,0.02016533414522807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,6144,0.026875555515289307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,6144,0.024661334024535284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,5120,0.017982222967677645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,5120,0.023119111855824787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,5120,0.02073955535888672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,4096,0.015448000695970325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,4096,0.01977511081430647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,4096,0.018047110901938546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,3584,0.014552889598740472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,3584,0.01731911136044396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,3584,0.016297777493794758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,3072,0.01310400002532535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,3072,0.015427556302812366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,3072,0.015390222271283468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,2560,0.011772444678677453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,2560,0.013441777891582914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,2560,0.013620444469981723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,2048,0.012495111260149213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,2048,0.011242667006121742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,1536,0.009268444445398118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,1536,0.009304888546466827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,1536,0.010904889139864178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,1024,0.007852444218264686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,1024,0.007242666350470648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,1024,0.009656888743241629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,768,0.006113777971929974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,768,0.006254222244024277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,768,0.008085333638721043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,2048,0.010408889088365767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,512,0.005471999860472149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,512,0.005094222310516569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,512,0.0069324444565508105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,256,0.0035297779573334586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,64,0.006244444598754247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,256,0.006518222391605377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,128,0.004652444273233414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,128,0.0032195556494924757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,128,0.006182222316662471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,64,0.003036444385846456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,8192,32,0.003129777808984121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,8192,32,0.006204444501135085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,16384,0.0450817777050866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,65536,0.2503395610385471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,16384,0.062202665540907115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,16384,0.03979466689957513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,65536,0.15615110927157932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,12288,0.03524355424775018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,12288,0.045224001010258995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,65536,0.15225688616434732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,12288,0.03360533383157518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,10240,0.02979644470744663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,10240,0.03881333271662394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,10240,0.029486222399605647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,8192,0.024872000018755596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,8192,0.03149155444569058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,8192,0.02325422234005398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,8192,256,0.004922666483455234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,7168,0.022608889473809138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,7168,0.027863999207814533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,7168,0.021486222743988037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,6144,0.019941333267423842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,6144,0.024542222420374554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,6144,0.019515555765893724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,5120,0.017438222964604694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,5120,0.02109866672092014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,5120,0.01742577718363868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,4096,0.016044444508022733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,3072,0.01423911088042789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,4096,0.017861333158281114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,3072,0.012906666431162091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,4096,0.014874666929244995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,3584,0.01422133379512363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,3584,0.015963556038008798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,3584,0.0138568886452251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,3072,0.012423111332787408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,2560,0.012409778104888068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,2560,0.011960888902346293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,2048,0.009849777652157677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,2048,0.010462221999963125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,2048,0.010856889188289642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,1536,0.008819555242856344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,1536,0.008741333252853816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,1536,0.009859555297427708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,1024,0.0070062221752272705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,1024,0.006858666737874349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,1024,0.008041777544551426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,768,0.005848000033034219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,768,0.005998222364319696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,768,0.006758222149478064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,512,0.00435288871328036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,512,0.006351111249791251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,2560,0.011095999843544431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,256,0.00480355570713679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,256,0.003359111025929451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,256,0.005888888819350137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,128,0.003088888815707631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,128,0.00565244464410676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,64,0.0029351111087534162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,64,0.005656888915432825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,7168,32,0.003003555453485913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,512,0.005471111171775394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,7168,32,0.005685333162546158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,16384,0.04611555404133267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,65536,0.21411289109124076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,65536,0.13505244255065918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,16384,0.056454221407572426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,16384,0.03771022293302748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,7168,128,0.004471111214823193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,12288,0.034344888395733304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,12288,0.04474577638838026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,12288,0.030213332838482324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,10240,0.029320889049106177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,10240,0.03484266665246751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,10240,0.026893332600593567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,8192,0.02404800057411194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,8192,0.02836266822285122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,8192,0.021998221675554912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,7168,0.02218666672706604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,65536,0.15897422366672093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,7168,0.025037333369255066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,7168,0.020622221959961783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,6144,0.0188080006175571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,6144,0.02216888964176178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,6144,0.018594667315483093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,5120,0.01683200067943997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,5120,0.019242667489581637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,5120,0.016661332713233102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,4096,0.014326221413082547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,4096,0.016391999191708036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,4096,0.014110222458839417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,3584,0.013433777623706393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,3584,0.01443555619981554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,3584,0.013295110729005603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,3072,0.012235555383894177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,3072,0.01276266657643848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,3072,0.01223288890388277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,2560,0.011001778145631155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,2560,0.011223999990357293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,2560,0.011527111132939657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,2048,0.009868444667922126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,2048,0.009824000298976898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,2048,0.010506666368908353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,1536,0.008647999829716152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,1536,0.008208889100286696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,1536,0.009526222116417354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,1024,0.006559999866618051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,1024,0.00648266656531228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,1024,0.00721066693464915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,768,0.00591555568906996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,768,0.005718222094906702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,768,0.006630222416586346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,512,0.005384888913896348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,512,0.003997333347797394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,512,0.006175999840100606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,256,0.0033235556135574975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,256,0.005821333163314396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,128,0.004562666846646203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,128,0.003051555612021022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,128,0.005603555589914322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,64,0.0028640000770489373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,64,0.005559111220969095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,6144,32,0.0029279999434947968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,6144,32,0.005592000153329637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,65536,0.18698044617970785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,65536,0.13490400049421522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,65536,0.1048284437921312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,16384,0.04322310951020983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,16384,0.04971555537647671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,6144,256,0.004937777916590373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,16384,0.031159109539455835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,12288,0.03241777751180861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,12288,0.03873422079616123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,12288,0.025220443805058796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,10240,0.027373333772023518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,10240,0.033391998873816595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,10240,0.022350221872329712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,8192,0.022838221655951604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,8192,0.026529777381155226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,8192,0.019077334139082167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,7168,0.02095199955834283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,7168,0.023916444844669763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,7168,0.01761866609255473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,6144,0.018420444594489205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,6144,0.021273778544531927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,6144,0.015987555185953777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,5120,0.016182222300105624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,5120,0.014676445060306124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,4096,0.014335110783576965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,4096,0.013741333451535968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,4096,0.012620444099108378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,3584,0.013017777767446307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,3584,0.01205244412024816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,3584,0.011921777493423887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,3072,0.012250666817029318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,3072,0.010999110837777456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,3072,0.011191999746693505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,2560,0.010616888602574667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,2560,0.009759110709031424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,2560,0.010457778142558204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,2048,0.009324444664849175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,2048,0.008482666479216682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,2048,0.009714666340086196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,5120,0.016199111938476562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,1536,0.008311111066076491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,1536,0.007367110914654202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,1536,0.008753777378135258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,1024,0.006155555446942647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,1024,0.005976000179847081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,1024,0.006745777610275481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,768,0.00573955559068256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,768,0.005056888692908817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,768,0.0063457778758472866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,512,0.0052577778697013855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,512,0.003679111186001036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,512,0.006001777946949005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,256,0.004713777866628435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,256,0.0032302221904198327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,256,0.005710222240951326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,5120,128,0.004383999854326248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,128,0.003012444410059187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,128,0.005478222337034013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,64,0.002809777855873108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,64,0.005514666852023866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,5120,32,0.002874666617976295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,5120,32,0.005583999885453119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,65536,0.13429778152041966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,65536,0.1388853391011556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,16384,0.040844443771574236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,12288,0.02184088859293196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,16384,0.039782222774293684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,65536,0.08627111381954616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,16384,0.027016000615225896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,12288,0.03003822101487054
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,12288,0.03130044539769491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,10240,0.026910222238964502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,10240,0.027075555589463975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,10240,0.01956355571746826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,8192,0.022710222336981032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,8192,0.01702844434314304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,7168,0.019714666737450492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,7168,0.02071644365787506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,7168,0.015475556254386902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,6144,0.017487110363112558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,6144,0.018537777993414137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,6144,0.014310222533014087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,5120,0.015449777245521545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,5120,0.01644888851377699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,5120,0.013239999612172445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,3584,0.012634666429625617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,4096,0.01332800007528729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,8192,0.02174844510025448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,4096,0.01386666629049513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,4096,0.011491555306646558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,3584,0.012479110724396177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,3584,0.010868444210953183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,3072,0.011821333732869891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,3072,0.011669333610269757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,3072,0.010303999814722273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,2560,0.010099555883142684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,2560,0.00867377801073922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,2560,0.009764444496896531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,2048,0.009105777574910058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,2048,0.00778311077091429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,2048,0.008954666554927826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,1536,0.006697777658700943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,1536,0.006990222467316522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,1024,0.006168889088763132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,1024,0.005524444497293896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,1024,0.00629066675901413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,768,0.003952888978852166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,768,0.006083555519580841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,512,0.005212444398138259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,512,0.003661333272854487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,512,0.005848000033034219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,256,0.004821333206362194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,256,0.003261333331465721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,256,0.005598222215970357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,128,0.00435644429590967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,128,0.0029493332323100832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,128,0.005435555759403441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,1536,0.00721066693464915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,64,0.003050666716363695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,64,0.0053582220441765254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,4096,32,0.003026666740576426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,4096,768,0.00563733321097162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,4096,32,0.005399111244413588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,65536,0.12581333849165174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,16384,0.038908445172839694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,65536,0.08154044548670451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,16384,0.03425511055522495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,16384,0.025222222010294598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,12288,0.028917332490285236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,12288,0.027256889475716486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,12288,0.020640000700950623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,10240,0.025240888198216755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,10240,0.023888889286253188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,10240,0.018522666560279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,8192,0.02052088909678989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,65536,0.13327377372317845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,8192,0.016099555624855887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,7168,0.019291554888089497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,7168,0.018499554859267343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,7168,0.014985778265529208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,6144,0.01739377776781718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,8192,0.021571555071406897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,6144,0.01661155621210734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,6144,0.013764444324705334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,5120,0.015007999208238391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,5120,0.0147386673423979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,5120,0.012734221915404001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,4096,0.014089778065681458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,4096,0.013079110946920184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,4096,0.011058666639857821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,3584,0.012175111307038201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,3584,0.012049777640236748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,3584,0.01055022246307797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,3072,0.011046222514576383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,3072,0.01107911103301578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,3072,0.010034666293197209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,2560,0.00979022185007731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,2560,0.00831733312871721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,2560,0.009549332989586724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,2048,0.008488000267081791
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,2048,0.007229333122571309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,2048,0.008400888906584846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,1536,0.006746666712893381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,1536,0.0064293332397937775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,1536,0.006879999819729064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,1024,0.0058622223635514574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,1024,0.005071111023426056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,1024,0.00629066675901413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,768,0.00546488869521353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,768,0.003943110919660992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,768,0.006048000107208888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,512,0.005066666752099991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,512,0.003576888806290097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,512,0.005852444304360284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,256,0.0047128887640105355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,256,0.0031893334041039148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,256,0.005574222240183089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3584,128,0.004412444515360726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,128,0.0029093333416514923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,128,0.005385777602593104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,64,0.002749333365095986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,64,0.005372444374693765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3584,32,0.0028284444577164124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3584,32,0.005426666802830166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,65536,0.1070551143752204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,65536,0.11997244093153213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,16384,0.030933333767784968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,16384,0.03404711021317376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,65536,0.07587733533647326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,16384,0.024324445260895625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,12288,0.037992000579833984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,12288,0.02464177707831065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,12288,0.019948444432682462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,10240,0.0248364441924625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,10240,0.02181688944498698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,10240,0.017940445078743827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,8192,0.020985777179400127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,8192,0.018753778603341844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,8192,0.015455111861228943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,7168,0.019197333190176222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,7168,0.016979555288950603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,7168,0.014334221680959066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,6144,0.017235555582576327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,6144,0.015254222684436373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,6144,0.013340444200568728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,5120,0.015097777048746744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,3584,0.011400889191362592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,5120,0.013759999639458127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,5120,0.012340444657537671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,4096,0.01223111069864697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,4096,0.010857777463065254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,3584,0.012157333393891653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,3584,0.0103493332862854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,3072,0.010612444745169746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,3072,0.010438222024175854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,3072,0.009932444327407414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,2560,0.009778666827413771
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,2560,0.007920888562997183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,2560,0.009009777671760982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,2048,0.007867555651399825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,2048,0.0070364442136552595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,2048,0.007267555428875818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,1536,0.006809777683681912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,1536,0.006274666637182236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,1536,0.006670222100284364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,1024,0.004528000122971005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,1024,0.006255110932721033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,4096,0.013403555585278405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,768,0.005464000006516774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,768,0.0039022221333450745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,768,0.006091555373536215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,512,0.004898666507667965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,512,0.003580444388919406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,512,0.005836444596449534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,256,0.004519999855094486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,256,0.0031555555760860443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,256,0.005542222410440445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,128,0.0041759999261962045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,128,0.0028986665937635633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,128,0.005390222287840313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,64,0.0027386666172080566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,64,0.005392888767851724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,3072,32,0.0028204443968004654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,3072,1024,0.0057795556883017225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,3072,32,0.005414222263627582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,65536,0.09481066465377808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,65536,0.10564177566104466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,16384,0.03268355462286208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,65536,0.0721564425362481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,16384,0.02758488886886173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,16384,0.023401778605249193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,12288,0.02681066592534383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,12288,0.0218240006102456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,12288,0.01907111042075687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,10240,0.01900533338387807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,10240,0.01718133356836107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,8192,0.016297777493794758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,8192,0.014888000157144336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,7168,0.01850488947497474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,7168,0.014741332994567024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,7168,0.013968888256284924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,6144,0.016420443852742512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,6144,0.013239111337396832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,6144,0.012979555461141797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,5120,0.012131555212868584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,5120,0.013893332746293811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,5120,0.01202044470442666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,4096,0.01259022206068039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,4096,0.010846222440401712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,4096,0.01054133309258355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,3584,0.011635555161370171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,3584,0.0101724440852801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,10240,0.02365600069363912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,3584,0.010283555421564314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,3072,0.010135999984211391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,3072,0.009548444714811113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,3072,0.009571555587980483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,2560,0.008610666626029545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,8192,0.02068533334467146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,2560,0.0075093333919843035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,2560,0.008092444803979661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,2048,0.007251555720965068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,2048,0.006781333436568578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,2048,0.0069884442620807225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,1536,0.006504000061088138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,1536,0.005802666561471091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,1536,0.006622222148709827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,1024,0.005673777725961473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,1024,0.004334222111437055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,1024,0.006234666539563074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,768,0.005200888961553574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,768,0.003891555385457145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,768,0.005988444305128521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,512,0.0035182221068276297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,512,0.0058044443527857465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,256,0.004464888738261329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,256,0.003142222141226133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,256,0.005477333234416113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,128,0.0028764444092909494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,128,0.005383111122581694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,64,0.0027662221756246355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,64,0.005312888986534542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2560,32,0.002809777855873108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,512,0.004832889056868023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2560,32,0.00536088893810908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,65536,0.1203991174697876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,65536,0.0749760005209181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,16384,0.03584711088074578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,65536,0.06938311126497057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,16384,0.023059555225902136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,16384,0.022450667288568284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,12288,0.024896888269318476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,12288,0.018826666805479262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,12288,0.018352889352374606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,10240,0.02422044508987003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,10240,0.01661422186427646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2560,128,0.0041742221348815495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,10240,0.016678222351604037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,8192,0.019536889261669584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,8192,0.014163555370436775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,8192,0.014470222923490735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,7168,0.013171555267439948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,6144,0.012229333321253458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,6144,0.015831111205948722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,4096,0.013020444247457715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,7168,0.013376000026861826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,6144,0.012693333129088083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,5120,0.011152000062995486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,5120,0.011855999628702799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,4096,0.009958222508430481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,4096,0.010194666683673859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,3584,0.011120888921949597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,3584,0.009183111290136972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,3584,0.009400000174840292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,3072,0.009460444251696268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,3072,0.008735111190213097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,7168,0.018114666144053142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,3072,0.008427555362383524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,2560,0.008445333275530074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,2560,0.007321777443091075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,2560,0.007343110938866933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,2048,0.007535999847782983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,2048,0.006529777828190062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,5120,0.014223999447292753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,2048,0.006866666591829724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,1536,0.005140444470776452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,1536,0.006714666469229593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,1536,0.0064479998416370815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,1024,0.005819555371999741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,1024,0.004159111115667555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,1024,0.006060444646411472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,768,0.005395555661784277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,768,0.0037973332736227247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,768,0.005922666854328579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,512,0.004947555561860402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,512,0.0034151110384199354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,512,0.005665777872006099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,256,0.004511111312442356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,256,0.003093333293994268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,256,0.005422222117582957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,2048,128,0.004396444393528832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,128,0.0028284444577164124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,128,0.0052844443255000645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,64,0.0026648888985315957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,64,0.005321777943107817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,2048,32,0.0027359999302360746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,2048,32,0.0052844443255000645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,65536,0.10285244385401408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,65536,0.055976000097062856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,16384,0.03435733252101474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,65536,0.06781599919001262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,16384,0.018783999813927543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,16384,0.02177511155605316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,12288,0.026727999250094097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,12288,0.01517777807182736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,12288,0.018059555027219985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,10240,0.013507555756303998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,10240,0.016334222422705758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,8192,0.02050488856103685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,8192,0.011990221838156382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,8192,0.014088888963063559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,7168,0.01830577850341797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,7168,0.011183111204041375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,7168,0.013191999660597907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,6144,0.01588622232278188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,6144,0.010275555153687796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,6144,0.012500444220172035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,5120,0.013979555832015144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,5120,0.009714666340086196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,5120,0.011502222054534487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,4096,0.011392888923486074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,4096,0.00904088881280687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,4096,0.008929777476522658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,3584,0.00871288859181934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,10240,0.02261688808600108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,3584,0.008403555386596257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,3072,0.009169778062237633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,3072,0.008258666429254744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,3072,0.007809778054555257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,2560,0.008044444852405125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,2560,0.006928000185224745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,2560,0.007377777662542131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,2048,0.00720711135201984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,2048,0.0057884446448749965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,2048,0.006882666713661618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,1536,0.006375999914275275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,1536,0.0047866664826869965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,1536,0.006540444576077991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,1024,0.005471111171775394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,1024,0.004102222207519743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,1024,0.006079999936951532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,768,0.005207111024194294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,768,0.0037911112109820047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,768,0.00591555568906996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,512,0.00479911102188958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,512,0.0034151110384199354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,512,0.00564177789621883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,256,0.004375111311674118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,256,0.003080000066094928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,256,0.0054560001525614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,128,0.0041164445380369825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,128,0.002846222163902389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,128,0.005295111073387994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1536,3584,0.010245333115259806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,64,0.0026791110220882627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,64,0.005305777821275923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1536,32,0.0027057776848475137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1536,32,0.005336888962321811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,65536,0.08198133442136976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,65536,0.04052177733845181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,65536,0.0668755571047465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,16384,0.023804444405767653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,16384,0.014474666780895658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,16384,0.021478222476111516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,12288,0.021486222743988037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,12288,0.012448888685968188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,12288,0.017677333619859483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,10240,0.011571555501884885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,10240,0.018796443939208984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,10240,0.016028443972269695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,8192,0.01661688917213016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,8192,0.010235555469989777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,8192,0.01365066650840971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,7168,0.009299555586444007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,7168,0.01240799989965227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,6144,0.013103110922707452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,6144,0.008578666382365758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,6144,0.01108266661564509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,5120,0.010248888697889116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,5120,0.008767999708652496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,5120,0.009646221995353699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,4096,0.010535111029942831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,4096,0.00795733349190818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,4096,0.00871377769443724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,3584,0.008975110948085785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,3584,0.007909333540333642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,3584,0.008117333054542542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,7168,0.014748444159825643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,3072,0.008467555873923832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,3072,0.007199111084143321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,3072,0.007719111111429002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,2560,0.006744888921578725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,2560,0.006703111032644908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,2560,0.007314666277832455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,2048,0.00620977787507905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,2048,0.005494222044944763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,2048,0.006868444383144379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,1536,0.005600888695981767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,1536,0.004780444420046276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,1536,0.006462222172154321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,1024,0.005008000052637524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,1024,0.004101333518822988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,1024,0.006078222145636876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,768,0.0037439999481042228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,768,0.005902222047249476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,512,0.003383999897374047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,512,0.005622222191757626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,256,0.0030782222747802734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,768,0.004911111046870549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,256,0.005412444472312927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,512,0.004603555632962121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,128,0.00406666679514779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,1024,256,0.004367111043797599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,128,0.0028355556229750314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,128,0.005290666802061929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,64,0.0026764445420768526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,64,0.0052871112194326185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,1024,32,0.002699555622206794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,1024,32,0.0052737775776121355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,65536,0.07157333029641046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,65536,0.031249778138266668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,65536,0.06639733579423693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,16384,0.02148088812828064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,16384,0.012111999922328524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,16384,0.021347555849287245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,12288,0.017860444055663217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,12288,0.01036622209681405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,12288,0.01755199995305803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,10240,0.015432000160217285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,10240,0.009837333526876237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,10240,0.01587199999226464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,8192,0.01236177815331353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,8192,0.009117333425415887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,8192,0.012625777886973487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,7168,0.00865955568022198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,7168,0.011342222491900126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,6144,0.010126222338941362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,6144,0.008188444707128737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,6144,0.010671999719407825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,5120,0.00887111077706019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,5120,0.0081386665503184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,5120,0.00961511085430781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,4096,0.00794577764140235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,4096,0.006994666324721442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,4096,0.008557333714432186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,3584,0.007456888755162557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,3584,0.00723555518521203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,3584,0.00813155538505978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,3072,0.006959999601046245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,3072,0.006646222124497096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,3072,0.007808888951937358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,2560,0.006462222172154321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,2560,0.006139555739031897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,2560,0.007333333293596904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,2048,0.006206222292449739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,7168,0.011076444553004371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,2048,0.005415110952324337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,2048,0.0068764446510208984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,1536,0.005640888793600931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,1536,0.0047786666287316214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,1536,0.006510222123728857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,1024,0.005076444397370021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,1024,0.004072000169091755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,1024,0.006121777825885349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,768,0.004765333400832282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,768,0.003681777666012446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,768,0.005879111174080107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,512,0.004559111264016894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,512,0.0033964444365766314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,512,0.005663110978073544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,256,0.004201777693298128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,256,0.003051555612021022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,256,0.005479111025730769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,768,128,0.004013333469629288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,128,0.0028106667515304354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,128,0.005296000176005893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,64,0.0026951111439201566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,64,0.0053119998839166425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,768,32,0.0026897777699761917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,768,32,0.0053226666318045715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,65536,0.06787111361821492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,65536,0.023408888114823237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,16384,0.020285333196322124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,65536,0.06624088684717815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,16384,0.010244444840484196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,16384,0.02123911182085673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,12288,0.009038222332795462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,12288,0.0158924443854226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,10240,0.013653332988421122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,10240,0.008808888494968414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,10240,0.013844444519943662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,8192,0.00833511104186376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,8192,0.012002666791280111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,7168,0.0107022225856781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,7168,0.007544000115659501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,7168,0.011193777951929303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,6144,0.009753777749008602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,5120,0.009552000297440423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,6144,0.0069822221994400024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,6144,0.01036533299419615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,5120,0.007123555574152205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,12288,0.015775110986497667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,4096,0.00794666674402025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,4096,0.006462222172154321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,8192,0.011681777735551199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,4096,0.00849422232972251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,3584,0.0074444446298811175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,3584,0.006979555719428592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,3584,0.008069333102968004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,3072,0.006958222223652734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,3072,0.006634666687912411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,5120,0.008887999587588841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,3072,0.007709333466158972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,2560,0.006092444476154115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,2560,0.007303111255168915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,2048,0.005985777825117111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,2048,0.005399999933110342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,2048,0.0068408888247278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,1536,0.0047484445903036325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,1536,0.006484444356626934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,1024,0.004973333328962326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,1024,0.004049777984619141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,1024,0.006050666587220297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,768,0.0037004442678557504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,768,0.005879111174080107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,512,0.0033688888781600525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,2560,0.0064444442590077715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,1536,0.00554577757914861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,512,0.005573333137565189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,768,0.0047582222355736625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,256,0.0041893331540955436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,256,0.003032888803217146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,256,0.005403555515739653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,128,0.00398488880859481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,128,0.0028275555620590844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,128,0.005225777626037598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,64,0.0026844443960322272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,64,0.005274666680230035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,512,512,0.004412444515360726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,512,32,0.0026808888134029177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,512,32,0.0052684446175893145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,65536,0.06821599933836195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,65536,0.015033778217103748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,65536,0.06600711080763075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,16384,0.008276444342401292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,16384,0.0196142229768965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,12288,0.01556622154182858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,12288,0.007128888534175024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,12288,0.015402666396564908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,10240,0.013445333474212222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,10240,0.006839111033413146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,10240,0.013744889034165276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,8192,0.011579555769761404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,8192,0.0068711112770769335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,8192,0.011881777809725868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,7168,0.01072177787621816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,7168,0.006675555474228329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,7168,0.011095999843544431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,16384,0.01924977699915568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,6144,0.009755555954244401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,6144,0.006500444478458828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,6144,0.010272888673676385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,5120,0.006865777903132969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,5120,0.009487999810112847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,4096,0.0064239998658498125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,4096,0.008408889174461365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,3584,0.00738666703303655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,3584,0.0069679998689227635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,3584,0.008042666647169325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,3072,0.006903111106819577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,3072,0.006559111177921295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,3072,0.007606222397751278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,5120,0.008806222014957005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,2560,0.006045333213276333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,2560,0.007251555720965068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,2048,0.005889777921968036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,2048,0.005426666802830166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,4096,0.007872000336647034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,2048,0.006746666712893381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,1536,0.005495999836259418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,1536,0.0047031111187405055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,1536,0.0063884444534778595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,1024,0.004943111290534337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,1024,0.004063111212518481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,1024,0.005978666659858491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,768,0.0046844445168972015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,768,0.003663111064169142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,768,0.0058017778727743365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,512,0.004416000097990036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,2560,0.006404444575309753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,512,0.003293333368168937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,512,0.005562666803598404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,256,0.0041511112617121804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,256,0.002965333354141977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,256,0.0053102220926019884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,256,128,0.003916444463862313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,128,0.0027644443843099806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,128,0.005230222311284807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,64,0.0026293332791990707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,64,0.005207111024194294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,256,32,0.002591111179855135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,256,32,0.005205333232879639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,65536,0.06821244292789035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,65536,0.01218577805492613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,65536,0.06528355677922566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,16384,0.019119999474949308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,16384,0.006631111105283101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,16384,0.018812444474962022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,12288,0.015536889433860779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,12288,0.006486222147941589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,12288,0.015332443846596612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,10240,0.01352266636159685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,10240,0.006629333313968446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,10240,0.013752000199423896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,8192,0.006493333313200209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,8192,0.011866666376590729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,7168,0.010729778144094678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,7168,0.0063368889192740125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,7168,0.011057777537239922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,6144,0.009684444301658207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,6144,0.006229333165619109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,6144,0.010279110736317104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,5120,0.008737777670224508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,5120,0.00657155571712388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,5120,0.009482666850090027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,4096,0.007841777470376756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,4096,0.006122666514582104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,4096,0.008390222158696916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,3584,0.007333333293596904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,3584,0.006590222318967183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,3584,0.00796977761718962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,3072,0.006816888848940532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,3072,0.00630044440428416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,3072,0.007622222105662028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,2560,0.0063440000845326324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,2560,0.005848000033034219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,2560,0.007213333414660559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,2048,0.0058613332609335584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,2048,0.005183111048407025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,2048,0.006743999818960826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,1536,0.006392000036107169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,1536,0.004504888835880491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,1536,0.005517333332035277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,1024,0.0048071112897661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,1024,0.005931555396980709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,1024,0.0038968887594011095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,768,0.004567999806669024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,768,0.0036480000449551474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,768,0.0057608890864584185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,512,0.004366222355100843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,512,0.0032426667296224167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,512,0.005492444253630108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,256,0.004120888809363048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,8192,0.0114702218108707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,256,0.0029155556112527847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,256,0.005316444569163852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,16,128,128,0.0038853333228164246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,128,0.0027315556589100095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,128,0.005151999907361137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,64,0.0025475554996066624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,64,0.005190222213665645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,128,32,0.0025840000145965153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,128,32,0.005196444276306364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,65536,0.00979466653532452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,65536,0.06332177586025663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,16384,0.006439111298984951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,16384,0.018778666853904724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,12288,0.006310222049554189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,12288,0.015305777390797934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,10240,0.006421333385838403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,10240,0.013678222066826291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,8192,0.006340444501903322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,8192,0.01183555523554484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,7168,0.006228444476922353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,7168,0.01108444482088089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,6144,0.006086222413513396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,6144,0.010248888697889116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,5120,0.006452444526884291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,5120,0.009511111511124505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,4096,0.00608711110221015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,4096,0.008387555678685507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,3584,0.006507555643717448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,3584,0.008009778128729926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,3072,0.006175999840100606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,3072,0.00759733302725686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,2560,0.00565244464410676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,1024,0.005976000179847081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,2560,0.0072266666425599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,2048,0.005061333378156026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,2048,0.006773333168692059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,1536,0.004452444612979889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,1536,0.006364444477690591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,1024,0.003802666647566689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,768,0.003499555504984326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,768,0.005787555542257097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,512,0.003198222153716617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,512,0.005531555662552516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,256,0.0028560000161329904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,256,0.0053048887186580235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,128,0.002696000039577484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,128,0.005142222262091107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,64,0.002528000002106031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,64,0.005134222408135732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,64,32,0.002554666664865282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,64,32,0.005180444568395615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,65536,0.008794666992293464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,65536,0.06176266405317518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,16384,0.006515555497672822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,16384,0.018730666902330186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,12288,0.0064560001095136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,12288,0.015260444747077094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,10240,0.006424888968467712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,10240,0.01370488852262497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,8192,0.006298666612969504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,8192,0.011791999969217511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,7168,0.006175999840100606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,7168,0.011085333095656501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,6144,0.00602311102880372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,6144,0.010207999911573198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,5120,0.006384000182151794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,5120,0.009485333330101436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,4096,0.005984000033802456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,4096,0.008374222450786168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,3584,0.006468444648716185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,3584,0.007996444072988298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,3072,0.006084444622198741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,3072,0.007606222397751278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,2560,0.005683555371231503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,2560,0.007224889265166388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,2048,0.005052444421582752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,2048,0.006748444504208035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,1536,0.004455111092991299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,1536,0.0063528890411059065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,1024,0.0037946667936113146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,1024,0.005912000106440649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,768,0.003510222252872255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,768,0.0057751110030545135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,512,0.0031831111345026228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,512,0.005526222288608551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,256,0.0028320000403457214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,256,0.0053191110491752625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,128,0.0027093332674768236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,128,0.005135111096832487
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,64,0.002559111143151919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,64,0.00517155561182234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,16,32,32,0.00254488881263468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,16,32,32,0.005174222091833751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,16384,0.5229857762654623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,16384,0.30244623290167916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,12288,0.39126311408148867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,10240,0.32752267519632977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,10240,0.17170311344994438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,12288,0.23358133104112414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,8192,0.262319114473131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,8192,0.14470933543311224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,10240,0.20119199487898085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,8192,0.15147199895646837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,7168,0.23105067676968047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,6144,0.19796355565388998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,7168,0.14344178305731878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,6144,0.10858488745159572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,6144,0.10909066597620647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,5120,0.16614754994710287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,5120,0.0914577775531345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,4096,0.07473155524995592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,5120,0.09315199984444512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,4096,0.13406311141120061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,4096,0.08368088801701863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,16384,0.26717156834072536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,3584,0.11661067273881699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,3584,0.06467911269929674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,3584,0.06710400183995564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,12288,0.203165332476298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,7168,0.11987733840942383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,3072,0.056109335687425405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,3072,0.1017342209815979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,3072,0.05988799863391452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,2560,0.0842204425070021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,2560,0.04740622308519152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,1536,0.031063109636306763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,2560,0.0506879985332489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,2048,0.04017511010169983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,2048,0.06827111376656426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,2048,0.04197333256403605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,1536,0.0525182220670912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,1536,0.033795555432637535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,1024,0.023031999667485554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,1024,0.03660533494419522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,1024,0.025560888979170058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,768,0.01992177797688378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,768,0.028816888729731243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,768,0.021798221601380244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,512,0.01529600057337019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,512,0.021946666969193354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,512,0.018410666121376883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,256,0.011720889144473605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,256,0.01128533316983117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,256,0.015941333439615037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,128,0.0075333333677715724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,128,0.013968888256284924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,64,0.005714666512277391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,64,0.011136000355084738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,65536,32,0.004397333496146732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,65536,32,0.011023110813564725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,65536,128,0.008825778133339351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,65536,0.2824515501658122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,16384,0.13790399498409694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,12288,0.056023114257388644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,16384,0.08509777651892768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,12288,0.10597422387864854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,65536,0.5628488858540853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,12288,0.06528000036875407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,65536,0.3445911142561171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,8192,0.0396728879875607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,10240,0.04750488864050972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,10240,0.0803075565232171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,10240,0.0559519992934333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,8192,0.07292711072497897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,8192,0.04588177800178528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,7168,0.056891553931766085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,7168,0.04120355513360765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,6144,0.031009776724709406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,6144,0.049293332629733615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,16384,0.07504266500473022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,6144,0.03675555520587497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,7168,0.035014222065607704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,5120,0.04199288951026069
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,5120,0.03177600105603536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,4096,0.02221333318286472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,4096,0.034585777256223894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,4096,0.026563555002212524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,3584,0.02067377832200792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,3584,0.03033777740266588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,3584,0.023952000670962866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,3072,0.018486221631368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,3072,0.02686133318477207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,3072,0.021571555071406897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,2560,0.016373333003785875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,2560,0.022965333527988855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,2560,0.018726223044925265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,5120,0.026763555076387193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,2048,0.01903377804491255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,2048,0.016605334149466622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,1536,0.012216000093354119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,1536,0.015238222148683337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,1536,0.014334221680959066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,2048,0.014263111684057446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,1024,0.010097777677906884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,1024,0.011027555498811932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,1024,0.012053333222866058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,768,0.00903911143541336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,768,0.009062222308582729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,768,0.010804444551467896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,512,0.007867555651399825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,512,0.007102222078376346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,512,0.009548444714811113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,256,0.0053484443989064955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,256,0.004881777697139316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,256,0.00720177756415473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,16384,128,0.004776000148720211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,128,0.0034675554682811103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,128,0.006879999819729064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,64,0.0032133333798911837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,64,0.006847111301289664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,16384,32,0.0031724443866146933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,16384,32,0.0069360000391801195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,16384,0.10541422499550714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,16384,0.0769875579410129
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,16384,0.06184533569547865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,65536,0.4413155449761285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,12288,0.05622933308283488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,65536,0.26495999760097927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,65536,0.24892711639404297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,12288,0.07972088787290785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,12288,0.04743555519315931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,10240,0.04821155468622843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,10240,0.07448355356852214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,10240,0.042035556501812406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,8192,0.03987555702527364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,8192,0.053724444574779935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,8192,0.034132444196277194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,7168,0.034754667017194964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,7168,0.04788177874353197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,7168,0.030786666605207656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,6144,0.0414720012081994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,6144,0.02791022260983785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,5120,0.02630399995379978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,5120,0.035050666994518705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,5120,0.024791111548741657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,4096,0.022192888789706763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,4096,0.028359111812379625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,4096,0.020444444484180875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,3584,0.02017866737312741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,3584,0.02531911101606157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,3584,0.018654222289721172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,3072,0.01772177716096242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,3072,0.022262222237057153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,3072,0.017134222719404433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,6144,0.030511998467975195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,2560,0.015592889653311836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,1536,0.01275199982855055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,2560,0.01908888916174571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,2560,0.015368888775507608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,2048,0.013408888545301227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,2048,0.016041777200169034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,2048,0.013702222042613559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,1536,0.011357333097192975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,1536,0.011922666596041786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,1024,0.009355555805895064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,1024,0.009398221969604492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,1024,0.010325333310498131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,768,0.008311111066076491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,768,0.007917332980367873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,768,0.009573332965373993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,512,0.0058959999846087555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,512,0.006270222365856171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,512,0.007752888732486301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,256,0.005160000175237656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,256,0.0038337777886125776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,256,0.006538666784763336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,128,0.0033413333197434745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,128,0.006244444598754247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,64,0.003036444385846456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,64,0.00620266670982043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,12288,32,0.0031306667046414483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,12288,32,0.006224888894293044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,16384,0.05775555637147692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,16384,0.09068622191747029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,65536,0.36826221148173016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,12288,128,0.0047111109726958806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,12288,0.06990400287840101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,65536,0.1989377737045288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,65536,0.22045332855648467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,16384,0.05589510997136434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,12288,0.04451466600100199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,12288,0.043064001533720225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,10240,0.059008002281188965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,10240,0.03843911157713996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,10240,0.03722933265897963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,8192,0.04848177896605598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,8192,0.030736890104081895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,7168,0.027965333726671007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,7168,0.0432933337158627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,7168,0.028265777561399672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,6144,0.02344177828894721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,6144,0.03806844353675842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,8192,0.03167644474241469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,6144,0.0251920007997089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,5120,0.02033955521053738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,5120,0.033301333586374916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,5120,0.02219999995496538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,4096,0.017320000463061862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,4096,0.02733777794573042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,4096,0.01866933372285631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,3584,0.016525333126386006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,2560,0.013627555635240344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,3584,0.021104888783560857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,3584,0.01715288890732659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,3072,0.01847555571132236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,3072,0.015744000673294067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,2560,0.0159582214223014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,2560,0.014416888356208801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,2048,0.013220444321632385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,2048,0.012717333104875354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,1536,0.010096888575288985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,3072,0.014557333456145393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,1536,0.010438222024175854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,1536,0.011272889044549731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,1024,0.008244444098737504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,1024,0.010001777774757808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,2048,0.011544889046086205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,768,0.007893333004580604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,768,0.007109333243634965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,768,0.009099555512269339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,512,0.005894222193294101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,512,0.005748444547255834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,512,0.007032000356250339
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,256,0.003532444437344869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,256,0.006511999915043513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,128,0.0047111109726958806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,128,0.0032515554792351196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,1024,0.008664000365469191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,128,0.006175111151403851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,64,0.002938666691382726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,10240,256,0.005117333183685939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,64,0.006182222316662471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,10240,32,0.002971555623743269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,10240,32,0.00620977787507905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,65536,0.2979084385765923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,16384,0.07473778062396579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,16384,0.04876177840762668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,16384,0.05305511090490553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,65536,0.18191377321879068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,12288,0.05778488847944471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,65536,0.1915280024210612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,12288,0.04380977816051907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,10240,0.048977779017554395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,10240,0.03417155477735732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,8192,0.039994666973749794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,8192,0.029845333761639063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,12288,0.03720266620318095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,10240,0.032637334532207914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,7168,0.02425066630045573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,8192,0.026456000076399908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,7168,0.03576888971858554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,7168,0.02684088879161411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,6144,0.020750222934616935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,6144,0.026591110560629103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,6144,0.024275556206703186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,5120,0.018434666925006442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,5120,0.02296177711751726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,5120,0.021457778082953557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,4096,0.01571733256181081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,4096,0.019359999232821994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,4096,0.01794577803876665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,3584,0.014909333652920194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,3584,0.017179555363125272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,3584,0.016255999604860943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,3072,0.013320888910028668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,3072,0.015189333094490899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,3072,0.015103111664454142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,2560,0.011954666839705573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,2560,0.013227555486891003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,2560,0.013878222141000958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,2048,0.010421333213647207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,2048,0.011075555450386472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,2048,0.012423111332787408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,1536,0.009229333036475712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,1536,0.009182222187519073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,1536,0.011002666420406766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,1024,0.00777422222826216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,1024,0.007168889045715332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,1024,0.009542221824328104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,768,0.006607999818192587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,768,0.006193777753247156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,768,0.007584888901975419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,512,0.005469333380460739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,512,0.005179555465777715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,512,0.00685777763525645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,256,0.004927111168702443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,256,0.003499555504984326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,256,0.006520888871616787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,8192,128,0.004475555486149258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,128,0.003140444349911478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,128,0.006146666490369373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,64,0.003018666679660479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,64,0.00610133343272739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,8192,32,0.003006222140457895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,8192,32,0.006080889039569431
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,65536,0.24469333224826387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,65536,0.16842222213745117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,16384,0.04862222075462341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,16384,0.06080977784262764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,65536,0.14890133010016546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,16384,0.03977510995335049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,12288,0.03569599986076355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,12288,0.04892533355289035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,12288,0.03149155444569058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,10240,0.03147911032040914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,10240,0.04068266683154636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,10240,0.027599111199378967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,8192,0.02540622154871623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,8192,0.03138400117556254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,8192,0.023376888699001733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,7168,0.02256711158487532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,7168,0.02778488728735182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,7168,0.021350221501456365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,6144,0.020304888486862183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,6144,0.024471999870406255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,6144,0.019285332825448778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,5120,0.01773244473669264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,5120,0.021128889587190416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,5120,0.017177777157889474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,4096,0.015503111812803479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,4096,0.017720000611411203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,4096,0.014860444598727755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,3584,0.014538667268223233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,3584,0.015784889459609985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,3584,0.013807999591032663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,3072,0.012538666526476542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,3072,0.01391377713945177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,3072,0.012900444368521372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,2560,0.011332444846630096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,2560,0.012134222520722283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,2560,0.011819555527634092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,2048,0.009977777798970541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,2048,0.010465777582592435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,2048,0.0107351111041175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,1536,0.00887822194231881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,1536,0.008649778034951951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,1536,0.009743111001120673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,1024,0.006839111033413146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,1024,0.00795377790927887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,768,0.006109333286682765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,768,0.005917333480384614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,768,0.006806222101052602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,512,0.00434044458799892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,512,0.006249777972698212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,256,0.0033333332588275275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,256,0.004795555439260271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,1024,0.007260444263617198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,256,0.0057955553962124725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,128,0.00433688900536961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,128,0.0030951110853089225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,128,0.005566222386227713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,64,0.0028657778683635923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,64,0.005537777725193236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,7168,32,0.0028826666788922418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,7168,512,0.00535999983549118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,7168,32,0.005497777627574072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,65536,0.21342755688561332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,65536,0.1617342233657837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,16384,0.048307554589377515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,65536,0.1339822212855021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,16384,0.0564862224790785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,16384,0.03891822364595201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,12288,0.03518133362134298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,12288,0.0398231115606096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,12288,0.03199022346072727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,10240,0.033047109842300415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,10240,0.034274667501449585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,10240,0.026378666361172993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,8192,0.02500266664557987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,8192,0.02804622385236952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,8192,0.02204088866710663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,7168,0.02277244461907281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,7168,0.02481422159406874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,7168,0.020212444994184706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,6144,0.02012266715367635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,6144,0.02193244463867611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,6144,0.01830400029818217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,5120,0.017128000656763714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,5120,0.01891377733813392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,5120,0.01667288939158122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,4096,0.014732443624072604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,4096,0.016371554798550077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,4096,0.0140346669488483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,3584,0.013920888304710388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,3584,0.014198222094111972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,3584,0.013193777865833707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,3072,0.012553777959611682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,3072,0.012669333153300814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,3072,0.012233778006500669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,2560,0.011331555744012197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,2560,0.011151110960377587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,2560,0.011200889117187925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,2048,0.010040000081062317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,2048,0.009806222385830348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,2048,0.010235555469989777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,1536,0.008775999976529015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,1536,0.00816888858874639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,1536,0.009373333719041612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,1024,0.00654666663871871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,1024,0.0064035554726918536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,1024,0.007093333535724216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,768,0.006222222414281633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,768,0.005592000153329637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,768,0.006559111177921295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,512,0.005392888767851724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,512,0.004007110993067424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,512,0.006147555592987273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,256,0.004924444274769889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,256,0.0032968889507982465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,256,0.0057297779454125305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,6144,128,0.004512000001139111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,128,0.003028444531891081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,128,0.005521777603361342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,64,0.0028195555011431375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,64,0.005481777919663324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,6144,32,0.002838222309947014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,6144,32,0.005500444521506627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,65536,0.1862471103668213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,65536,0.14789511097802058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,16384,0.045626666810777455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,16384,0.04896533489227295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,65536,0.10433244705200195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,16384,0.031488001346588135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,12288,0.03347288899951511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,12288,0.0380364457766215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,12288,0.026791110634803772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,10240,0.028395556741290625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,10240,0.032580445210138954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,10240,0.022392888863881428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,8192,0.023299554983774822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,8192,0.02721066607369317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,8192,0.018952000472280715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,7168,0.022122666239738464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,7168,0.02476177778508928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,7168,0.017518222332000732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,6144,0.019325332509146798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,6144,0.020983111527231004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,6144,0.01573511130279965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,5120,0.016733333468437195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,5120,0.01848355597919888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,5120,0.014559111661381192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,4096,0.014194667339324951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,4096,0.015835555063353646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,4096,0.012617777619096967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,3584,0.013423110875818463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,3584,0.011976888610257043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,3584,0.011916444533401065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,3072,0.012044444680213928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,3072,0.010948444406191507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,3072,0.011193777951929303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,2560,0.010750222537252637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,2560,0.009676444033781687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,1536,0.0087333329849773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,2560,0.010407999985747868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,2048,0.009311111436949836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,2048,0.008456889126035903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,2048,0.009581333233250512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,1536,0.008396444221337637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,1536,0.0073520003093613525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,1024,0.006337777607970768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,1024,0.005950222412745158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,1024,0.006694222076071634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,768,0.006011555592219035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,768,0.005005333158704969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,768,0.006340444501903322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,512,0.005221333354711533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,512,0.003621333175235324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,512,0.006024000131421619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,256,0.0046693334976832075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,256,0.0032293332947625052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,256,0.005608888963858287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,5120,128,0.004376888689067629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,128,0.00300177786913183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,128,0.00543200017677413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,64,0.002768888862596618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,64,0.005438222239414851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,5120,32,0.0027671110712819626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,5120,32,0.005419555637571547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,65536,0.15084889199998644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,16384,0.04225511021084256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,65536,0.15143822299109563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,65536,0.08646221955617268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,16384,0.03963199920124478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,16384,0.026735999517970618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,12288,0.030902223454581365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,12288,0.031135112047195435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,12288,0.021512000097168818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,10240,0.026740445031060114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,10240,0.02678577767478095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,10240,0.019300444258583915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,8192,0.022587555978033278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,8192,0.022436444958051045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,8192,0.016519111063745286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,7168,0.02049066623051961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,7168,0.020340444313155282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,7168,0.01531644496652815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,6144,0.018361777067184448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,6144,0.014158222410413953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,5120,0.016040000650617812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,5120,0.0162435554795795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,5120,0.013013333082199097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,6144,0.017865777015686035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,4096,0.013604444762070974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,4096,0.01401511165830824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,4096,0.011340444286664328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,3584,0.01276000009642707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,3584,0.012461333639091916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,3584,0.01072266697883606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,3072,0.011518222590287527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,3072,0.011438222395049201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,3072,0.010253333383136326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,2560,0.01036711119943195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,2560,0.008450667063395182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,2560,0.00961688905954361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,2048,0.009223110973834991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,2048,0.0075502221783002215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,2048,0.008829333715968663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,1536,0.007322666545708974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,1536,0.00655644428398874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,1536,0.006828444285525217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,1024,0.005673777725961473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,1024,0.00627377794848548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,768,0.006075555665625467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,768,0.0041680000722408295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,768,0.005994666781690385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,512,0.0052062223354975385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,512,0.0035537779331207275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,512,0.005752000129885144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,256,0.0047271110945277745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,1024,0.006146666490369373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,256,0.003170666595300039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,256,0.00554933316177792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,4096,128,0.004368888835112254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,128,0.002862222285734283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,128,0.005391110976537068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,64,0.0027235555979940626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,64,0.005277333160241445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,4096,32,0.002765333279967308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,4096,32,0.005306666509972678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,16384,0.040511111418406166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,16384,0.03582488828235202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,65536,0.12346755133734809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,65536,0.07944355408350627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,16384,0.025018667181332905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,12288,0.029463112354278564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,12288,0.026776888304286536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,12288,0.02036977807680766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,10240,0.028668443361918133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,10240,0.02297244469324748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,10240,0.01827822294500139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,8192,0.0218248897128635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,8192,0.02034133341577318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,8192,0.01594311164485084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,7168,0.02016266683737437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,7168,0.01811288959450192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,7168,0.015007999208238391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,6144,0.017814222309324477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,6144,0.01633155511485206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,6144,0.013569778038395775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,5120,0.015385778413878547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,5120,0.014491554763582019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,5120,0.012507555385430654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,4096,0.013212444053755866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,4096,0.012764444781674279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,4096,0.010953777366214328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,65536,0.14099733034769693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,3584,0.0124177775449223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,3584,0.01182933317290412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,3584,0.010431999961535135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,3072,0.011271110839313932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,3072,0.01086400035354826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,3072,0.009948444035318162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,2560,0.009983110758993361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,2560,0.008153777983453538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,2560,0.009335999687512716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,2048,0.008558221989207799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,2048,0.007133333219422235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,2048,0.008182221816645728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,1536,0.006872889068391588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,1536,0.00637155564294921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,1536,0.0069422221018208405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,1024,0.005896889087226655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,1024,0.0047377778424157035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,1024,0.006227555374304454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,768,0.005712888720962737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,768,0.0038542221817705366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,768,0.00600888869828648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,512,0.005146666533417172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,512,0.003522666792074839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,512,0.005783111270931032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,256,0.0046942221621672315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,256,0.003175111073586676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,256,0.005499555418888728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,128,0.002874666617976295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,128,0.0052720002002186244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,64,0.002696000039577484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,64,0.005494222044944763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3584,32,0.0027466666781240036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3584,32,0.005316444569163852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,65536,0.1075964437590705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,65536,0.11949777603149414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,65536,0.07558133204778035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,16384,0.030388444662094116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,16384,0.023966223001480103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,12288,0.030518223841985066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,12288,0.02405333353413476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3584,128,0.004211555752489301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,12288,0.01943733294804891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,10240,0.02532800038655599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,10240,0.021337777376174927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,10240,0.017673777209387887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,16384,0.03446311089727614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,8192,0.018420444594489205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,8192,0.015171556009186639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,7168,0.01959822244114346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,7168,0.016694222887357075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,7168,0.014124444789356656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,6144,0.017112889223628573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,6144,0.014975999792416891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,6144,0.013176889055305056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,5120,0.015419556034935845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,5120,0.01349955548842748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,5120,0.012272000312805176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,4096,0.013409777647919126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,4096,0.012080888781282636
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,4096,0.010751999914646149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,3584,0.012614222036467658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,3072,0.009665778113736046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,3584,0.011145778000354767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,3584,0.010403555300500657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,3072,0.011253332926167382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,3072,0.010319111247857412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,2560,0.00962577760219574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,2560,0.007814221911960179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,2560,0.008728889127572378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,2048,0.007648000286685095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,2048,0.00684533350997501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,2048,0.007279111279381647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,1536,0.007164444360468123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,1536,0.006191111273235745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,1536,0.006583999842405319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,8192,0.021756443712446425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,1024,0.004583999928500918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,1024,0.006215111249023014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,768,0.005730666634109285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,768,0.003830222205983268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,768,0.005971555494599872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,512,0.004998222407367495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,512,0.003506666670242945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,512,0.00573244442542394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,256,0.00453333349691497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,256,0.0031804444475306403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,256,0.005497777627574072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,128,0.0041511112617121804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,128,0.0028640000770489373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,128,0.005286222116814719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,64,0.0027066665805048416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,3072,1024,0.0058604445722368025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,64,0.0052977779673205475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,3072,32,0.0027253333893087176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,3072,32,0.005296000176005893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,65536,0.09570133023791844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,65536,0.10818222496244644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,65536,0.07151555352740817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,16384,0.038867556386523776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,16384,0.026943999860021803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,16384,0.023039110832744177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,12288,0.02873599860403273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,12288,0.02147644427087572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,12288,0.018791110979186166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,10240,0.021147555775112573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,10240,0.018548443913459778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,10240,0.0169831116994222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,8192,0.02184355590078566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,8192,0.01606666710641649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,8192,0.014696000350846184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,7168,0.01832088828086853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,7168,0.014701333310869006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,7168,0.013749333719412485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,6144,0.017255110873116385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,6144,0.013152889079517789
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,6144,0.012753778033786349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,5120,0.015032889114485847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,5120,0.011983111500740051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,5120,0.0119982221060329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,4096,0.01295377810796102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,4096,0.010739555789364709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,4096,0.01036177741156684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,3584,0.011958222422334882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,3584,0.0101742222905159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,3584,0.010078222387366826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,3072,0.010599999792046018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,3072,0.009488888912730748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,3072,0.009453333086437648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,2560,0.008025777836640676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,2560,0.007401777638329401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,2560,0.007511999871995714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,2048,0.007502222226725684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,2048,0.006673777682913675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,2048,0.0069724445541699724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,1536,0.006708444406588872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,1536,0.005612444546487596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,1536,0.006528888725572162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,1024,0.005621333503060871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,1024,0.0042275554604000514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,1024,0.006133333262470033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,768,0.005506666584147348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,768,0.0038399998512532976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,768,0.005932444499598608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,512,0.004815111143721474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,512,0.0034719999465677473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,512,0.005684444473849402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,256,0.004430222428507275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,256,0.0031031111462248694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,256,0.00543200017677413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2560,128,0.004081777814361784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,128,0.0028453332682450614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,128,0.0052586665583981415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,64,0.0026897777699761917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,64,0.005249777601824866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2560,32,0.0027235555979940626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2560,32,0.005215999980767568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,65536,0.1357591152191162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,65536,0.07402755816777547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,65536,0.06909688976075914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,16384,0.022646221849653456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,10240,0.024482667446136475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,16384,0.0221404449807273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,12288,0.027757333384619817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,12288,0.018422222799725003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,12288,0.018173333671357896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,10240,0.01626311077011956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,8192,0.02091022166940901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,10240,0.01648355523745219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,8192,0.013975110318925647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,8192,0.01421866648726993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,7168,0.018479110466109384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,7168,0.012977778083748288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,7168,0.013356444736321768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,6144,0.016349333855840895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,6144,0.01201600001917945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,6144,0.012478222449620565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,4096,0.010152888794740042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,5120,0.014671110444598727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,5120,0.010911111202504901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,5120,0.011612444288200803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,4096,0.012267555627557965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,4096,0.009856888817416297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,3584,0.011223999990357293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,3584,0.009288888838556077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,16384,0.038507554266187884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,3584,0.009125333693292405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,3072,0.009751111268997192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,3072,0.008731555607583787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,3072,0.007687110867765214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,2560,0.00869777798652649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,2560,0.007231111327807109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,2560,0.007294221884674496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,1024,0.005797333187527127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,2048,0.006437333507670297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,2048,0.00833422193924586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,2048,0.006789333290523953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,1536,0.006868444383144379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,1536,0.005192888693677055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,1536,0.006432000133726332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,1024,0.004060444318585926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,1024,0.005991999887757831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,768,0.005774222314357758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,768,0.0037457777394188773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,768,0.0057848890622456866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,512,0.005031110925806893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,512,0.0033893332713180115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,512,0.005588444570700328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,256,0.004658666749795278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,256,0.0030133333057165146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,256,0.00535022219022115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,2048,128,0.004233777936961916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,128,0.002768888862596618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,128,0.005179555465777715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,64,0.0026444445053736367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,64,0.0052133335007561584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,2048,32,0.002687999978661537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,2048,32,0.0051857779423395796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,65536,0.11094933085971409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,65536,0.05577066871854994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,16384,0.018328888548745047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,65536,0.06707733207278781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,16384,0.03658399979273478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,16384,0.021643555826610986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,12288,0.026642666922675237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,12288,0.01479111115137736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,12288,0.01779733267095354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,10240,0.023570666710535686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,10240,0.013280889226330651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,10240,0.016172443826993305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,8192,0.021036444438828364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,8192,0.0118231111102634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,8192,0.014016888207859464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,7168,0.018505778577592637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,7168,0.010978666444619497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,7168,0.013117333253224691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,6144,0.01584533353646596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,6144,0.01015644437736935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,6144,0.012304888831244575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,5120,0.013714666995737286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,5120,0.00942488925324546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,5120,0.011360000405046674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,4096,0.011364444262451597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,4096,0.00870311094654931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,4096,0.008809777597586313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,3584,0.010503110786279043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,3584,0.00850399997499254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,3584,0.008303999900817871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,3072,0.009233777721722921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,3072,0.00816177742348777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,3072,0.007677333222495184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,2560,0.00813066628244188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,2560,0.0069102222720781965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,2560,0.007286222444640265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,2048,0.007359999749395583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,2048,0.00544088871942626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,2048,0.006798222247097228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,1536,0.006502222269773483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,1536,0.004734222259786394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,1536,0.006454222318198945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,1024,0.005592000153329637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,1024,0.004053333153327306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,1024,0.006060444646411472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,768,0.005400889035728242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,768,0.003657777690225177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,768,0.0058311112225055695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,512,0.004835555536879433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,512,0.0034008889148632684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,512,0.005575110928879843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,256,0.004376888689067629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,256,0.003050666716363695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,256,0.005408000200986862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1536,128,0.003999999827808804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,128,0.0028008888992998335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,128,0.005240889059172736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,64,0.0026577777332729767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,64,0.005220444252093633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1536,32,0.002675555646419525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1536,32,0.005285333428117964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,65536,0.03968977928161621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,65536,0.0888026687833998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,65536,0.06635822190178765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,16384,0.029072887367672388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,16384,0.014222222897741528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,16384,0.021319111188252766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,12288,0.025215110845035974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,12288,0.012257777982287936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,12288,0.017656000124083627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,10240,0.01798311041461097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,10240,0.01124622258875105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,10240,0.01587466730011834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,8192,0.018238221605618794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,8192,0.010046222143703038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,8192,0.013416888813177744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,7168,0.01482488877243466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,7168,0.009334222310119206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,7168,0.01275111072593265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,6144,0.013800000150998434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,6144,0.008431110945012834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,6144,0.01054666688044866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,5120,0.011927111281288994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,5120,0.008461332983440822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,5120,0.00960177762640847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,4096,0.00790666706032223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,4096,0.008492444124486711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,3584,0.009424000150627559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,3584,0.007606222397751278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,4096,0.010221333139472539
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,3584,0.008135110967689091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,3072,0.008753777378135258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,3072,0.0069208890199661255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,3072,0.007655111451943715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,2560,0.0069155556460221606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,2560,0.006144888699054718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,2560,0.007279111279381647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,2048,0.00646311127477222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,2048,0.005416000054942236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,2048,0.006783111227883234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,1536,0.0058853332367208265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,1536,0.004752000172932942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,1536,0.006396444307433234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,1024,0.00507377791735861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,1024,0.004045333299371931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,1024,0.00600888869828648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,768,0.005184000151024925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,768,0.003678222083383136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,256,0.00535733335547977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,768,0.005807999935415056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,512,0.0046657779150538975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,512,0.0033982222278912864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,512,0.005575110928879843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,256,0.004357333398527569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,256,0.0030382221771611106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,128,0.0028106667515304354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,1024,128,0.004041777716742622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,128,0.005196444276306364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,64,0.0026773332307736077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,64,0.005202666752868229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,1024,32,0.0026773332307736077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,1024,32,0.005124444348944558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,65536,0.08067022429572211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,65536,0.03059555424584283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,65536,0.0658444431093004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,16384,0.023952888117896184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,16384,0.01186044431395001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,16384,0.02129688858985901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,12288,0.010178666975763109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,12288,0.01759288873937395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,10240,0.0170133329100079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,10240,0.009750222166379293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,10240,0.01587288909488254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,8192,0.013551111022631327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,12288,0.020084444019529555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,8192,0.008960000342792934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,8192,0.012814222110642327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,7168,0.013039111263222165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,7168,0.008447111480765873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,7168,0.011459555890825061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,6144,0.00980088859796524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,6144,0.007968000239796108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,6144,0.01037333326207267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,5120,0.009766221874290043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,5120,0.00795022232664956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,5120,0.009577777650621202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,4096,0.008276444342401292
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,4096,0.00684799998998642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,4096,0.008444444172912175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,3584,0.007763555480374231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,3584,0.006978666616810693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,3584,0.008059555457697975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,3072,0.007105777661005656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,3072,0.006771555377377405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,3072,0.007698666718271043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,2560,0.006483555667930179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,2560,0.008184000021881526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,2560,0.007273777491516537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,2048,0.005910222315125995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,2048,0.007015110717879401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,2048,0.006818666640255187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,1536,0.005525333185990651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,1536,0.005923555543025334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,1536,0.006446222050322427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,1024,0.004922666483455234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,1024,0.004634666774008009
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,1024,0.005990222096443176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,768,0.0047973332305749255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,768,0.004271999829345279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,768,0.0058417779703934984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,512,0.004511111312442356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,512,0.003730666720204883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,512,0.005588444570700328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,256,0.0041928887367248535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,256,0.003216000066863166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,256,0.005376889059940974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,768,128,0.003958222352796131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,128,0.0029688889367712867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,128,0.0052266667286554975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,64,0.0028284444577164124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,64,0.005247999810510212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,768,32,0.0028124445428450904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,768,32,0.005151111218664381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,65536,0.0681697792477078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,65536,0.022866666316986084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,65536,0.06577955351935492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,16384,0.020222221811612446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,16384,0.009904888768990835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,16384,0.021088000800874498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,12288,0.015663999650213454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,12288,0.008797333472304875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,12288,0.015574221809705099
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,10240,0.013631111217869652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,10240,0.008640000389681922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,10240,0.013712888790501488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,8192,0.00815822184085846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,8192,0.011898666620254517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,7168,0.010750222537252637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,7168,0.007243555453088548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,7168,0.011155555645624796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,6144,0.00675733346078131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,6144,0.010301333334710863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,5120,0.00886577781703737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,5120,0.006855999843941794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,8192,0.011696000066068439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,5120,0.009530666801664565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,6144,0.00979288915793101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,4096,0.007868444754017724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,4096,0.00648266656531228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,4096,0.008426666259765625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,3584,0.0074417781498697065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,3584,0.0069617778062820435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,3584,0.008084444536103142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,3072,0.006976000136799282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,3072,0.006627555522653792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,3072,0.007677333222495184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,2560,0.006451555424266391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,2560,0.006197333335876465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,2560,0.007253333098358578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,2048,0.005524444497293896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,2048,0.006796444455782573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,1536,0.005509333478079901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,1536,0.004870222260554631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,768,0.003793777690993415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,1536,0.006416000011894438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,1024,0.004888888862397936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,1024,0.004138666722509596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,1024,0.006000888844331105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,768,0.0046471113132105935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,256,0.0031519999934567344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,768,0.005767111149099138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,512,0.004368000146415499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,512,0.0034835553831524322
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,512,0.005564444594913059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,256,0.004209777961174647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,256,0.005311111195219888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,128,0.0039013334446483185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,128,0.0028595555987623003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,128,0.00519466648499171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,64,0.0027431110954946945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,64,0.0051857779423395796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,512,32,0.0027066665805048416
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,512,32,0.00518488883972168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,512,2048,0.0058906666106647914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,65536,0.014667555689811707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,65536,0.0651351081000434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,16384,0.019103111492262948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,16384,0.008085333638721043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,16384,0.01885955532391866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,12288,0.015495111544926962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,12288,0.006892444358931647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,12288,0.015248000621795654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,10240,0.013442666994200813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,10240,0.006833777659469181
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,10240,0.013565333353148567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,8192,0.01162755572133594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,8192,0.006586666736337874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,8192,0.011760888828171624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,7168,0.006589333216349284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,7168,0.011017777853541903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,65536,0.0676800012588501
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,6144,0.009720889230569204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,6144,0.006355555521117316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,7168,0.010737777584128909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,6144,0.010245333115259806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,5120,0.008773333496517604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,5120,0.006769777586062749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,5120,0.009450666606426239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,4096,0.006341333190600078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,4096,0.008349333372380998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,3584,0.007332444190979004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,3584,0.006832888970772426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,3584,0.00792622235086229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,3072,0.006877333339717653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,3072,0.006474666711356904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,3072,0.00755466686354743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,2560,0.0063440000845326324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,2560,0.006064888917737537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,2560,0.007186666958861881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,2048,0.005869333528810077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,2048,0.005361777626805835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,4096,0.007836444510353936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,2048,0.006728888799746831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,1536,0.005430222385459476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,1536,0.004693333473470476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,1536,0.006270222365856171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,1024,0.004879999905824661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,1024,0.004004444513056013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,1024,0.0058746664888328975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,768,0.004678222040335338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,768,0.003671110918124517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,768,0.005671111245950063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,512,0.004376000000370873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,512,0.0033119999700122406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,512,0.00545866663257281
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,256,0.004098666624890434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,256,0.003006222140457895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,256,0.005260444349712796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,128,0.0027724444452259275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,128,0.005070222334729301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,64,0.0026293332791990707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,64,0.005113777601056629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,256,32,0.0026168889469570587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,256,32,0.005131555514203178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,65536,0.06694310903549194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,65536,0.011503111157152386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,65536,0.0635662211312188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,16384,0.019137778215938143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,16384,0.0064071110553211635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,16384,0.01863822175396813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,12288,0.006305777778228124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,12288,0.015115555789735583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,10240,0.01348977784315745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,10240,0.006416000011894438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,10240,0.013624889155228933
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,8192,0.011585777832402123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,8192,0.0064177778032090925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,8192,0.011698666546079846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,7168,0.010590222146775989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,7168,0.006231110956933763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,7168,0.010975110861990186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,6144,0.009665778113736046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,6144,0.00611200018061532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,6144,0.010127999716334874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,5120,0.008741333252853816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,256,128,0.0037911112109820047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,5120,0.006405333264006509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,5120,0.009400889277458191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,4096,0.00777511133088006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,4096,0.006047111004590988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,4096,0.008270222279760573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,3584,0.007289778027269576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,3584,0.006500444478458828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,3584,0.007917332980367873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,3072,0.006948444578382704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,3072,0.006234666539563074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,3072,0.0075297777851422625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,2560,0.0063475556671619415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,12288,0.015425778097576566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,2560,0.006060444646411472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,2560,0.007129777636792924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,2048,0.0057848890622456866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,2048,0.005395555661784277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,2048,0.006671111202902264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,1536,0.0046657779150538975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,1536,0.006285333385070165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,1024,0.00479911102188958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,1024,0.004032000071472592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,1024,0.0058408888677755995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,768,0.004578666554556953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,768,0.003628444340493944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,768,0.005670222143332164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,512,0.004278222305907143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,512,0.003314666656984223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,512,0.005434666656785541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,256,0.004036444342798657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,256,0.003002666764789157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,256,0.0052133335007561584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,128,0.0037004442678557504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,128,0.0027306667632526825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,8,128,1536,0.005355555564165115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,128,0.005117333183685939
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,64,0.0025742221623659134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,64,0.005118222286303838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,128,32,0.0025724443710512584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,128,32,0.005049777941571342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,65536,0.009296000003814697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,65536,0.06223022275500827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,16384,0.006263111200597551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,16384,0.018650665879249573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,12288,0.006166222194830577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,12288,0.01514755520555708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,10240,0.0063982220987478895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,10240,0.013605333036846586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,8192,0.006262222097979651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,8192,0.011703111231327057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,7168,0.006140444427728653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,7168,0.010969777901967367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,6144,0.005973333285914526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,6144,0.01014666673209932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,5120,0.00638488887084855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,5120,0.009375111096435124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,4096,0.00598933340774642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,4096,0.008293333152929941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,3584,0.0064080001579390625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,3584,0.00794933322403166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,3072,0.006164444403515921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,3072,0.007516444557242923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,2560,0.00565244464410676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,2560,0.007244444555706448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,2048,0.00499555551343494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,2048,0.00664533343580034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,1536,0.004367111043797599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,1536,0.006263111200597551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,1024,0.0037671112351947357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,1024,0.005856888989607493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,768,0.0034737777378824023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,768,0.005626666463083691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,512,0.003140444349911478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,512,0.005437333136796951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,256,0.0028479999552170434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,256,0.0052062223354975385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,128,0.0026568888376156488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,128,0.005114666703674528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,64,0.002540444541308615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,64,0.005037333402368757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,64,32,0.0025048889219760895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,64,32,0.005037333402368757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,65536,0.008593777815500895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,65536,0.059768888685438365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,16384,0.006274666637182236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,16384,0.018610666195551556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,12288,0.0061928890645504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,12288,0.015188443991872998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,10240,0.006292444550328785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,10240,0.013521778086821238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,5120,0.006325333482689328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,8192,0.006175999840100606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,8192,0.011769777370823754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,7168,0.006099555641412735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,7168,0.010974222587214576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,6144,0.005934222290913264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,6144,0.01015822258260515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,5120,0.009373333719041612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,4096,0.005949333310127258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,4096,0.008271999657154083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,3584,0.006385777973466449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,2048,0.006621333460013072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,3584,0.007883555359310573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,3072,0.006082666830884085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,3072,0.0075128889746136135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,2560,0.005955555372767978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,2560,0.007085333267847697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,2048,0.0052684446175893145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,1536,0.0046675557063685525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,1536,0.006238222122192383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,1024,0.00378222225440873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,1024,0.005818666683302985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,768,0.0034577778230110803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,768,0.005663110978073544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,512,0.0032044444233179092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,512,0.005428444594144821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,256,0.0028213332924577924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,256,0.005186666631036335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,128,0.0026515554636716843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,128,0.005038222091065513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,64,0.0024906665914588505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,64,0.005036444299750858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,8,32,32,0.002504000026318762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,8,32,32,0.005056888692908817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,16384,0.5232835345798069
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,12288,0.3914782206217448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,10240,0.3268391026390923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,12288,0.216392887963189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,16384,0.2808462248908149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,10240,0.17310133245256212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,12288,0.23971109920077852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,16384,0.30054132143656415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,8192,0.2616186671786838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,8192,0.14582578341166177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,10240,0.19957955678304037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,8192,0.14637777540418836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,7168,0.12436266740163167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,7168,0.22968000835842559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,6144,0.19738399982452393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,6144,0.11362577809227838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,7168,0.1348631117078993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,5120,0.16597778267330593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,5120,0.09392088651657104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,6144,0.12108177608913845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,4096,0.07766666677263048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,5120,0.094778663582272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,4096,0.13340177800920275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,3584,0.11617778407202826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,4096,0.07574133078257243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,3584,0.06984977589713202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,3584,0.07110933462778728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,3072,0.10085155566533406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,3072,0.05972266859478421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,3072,0.060552888446384005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,2560,0.08443644311692979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,2560,0.05191911260286967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,2560,0.052128000391854175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,2048,0.04116355710559421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,2048,0.06819999880260892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,2048,0.0420959989229838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,1536,0.03564800156487359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,1536,0.05212355653444926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,1536,0.03652711047066583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,1024,0.02349333299530877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,1024,0.0363377763165368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,1024,0.02535999980237749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,768,0.020799110333124798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,768,0.02791822287771437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,768,0.021939555803934734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,512,0.015127110812399121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,128,0.008784000244405534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,512,0.01997244523631202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,512,0.018204443984561496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,65536,256,0.01163111130396525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,256,0.011370666325092316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,256,0.015923556354310777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,128,0.00739644467830658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,128,0.013859555953078799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,64,0.00544088871942626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,64,0.010986666712496014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,65536,32,0.004323555363549126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,65536,32,0.011005333728260465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,65536,0.2843688858879937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,65536,0.562172465854221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,65536,0.342920011944241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,16384,0.13745066854688856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,16384,0.07597244448131986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,16384,0.08462666802936131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,12288,0.056792888376447886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,12288,0.09468799829483032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,12288,0.07003288798862033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,10240,0.05136355426576403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,8192,0.04619200030962626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,10240,0.09060088793436687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,10240,0.05609777900907728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,8192,0.03993599944644504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,8192,0.06441599792904325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,7168,0.03605688942803277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,7168,0.05690933598412407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,7168,0.0447653333346049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,6144,0.03087644444571601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,6144,0.04936000042491489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,6144,0.036068442795011736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,5120,0.02693066663212246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,5120,0.04195555713441637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,5120,0.03181155522664388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,4096,0.02314044369591607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,4096,0.03468088971243964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,4096,0.025983999172846477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,3584,0.022457778453826904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,3584,0.0302746660179562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,3584,0.024159111910396155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,3072,0.018477333916558158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,3072,0.026592888765864905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,3072,0.021563554803530376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,2560,0.018351111147138808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,2560,0.022838221655951604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,2560,0.019277332557572257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,2048,0.014156444205178155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,2048,0.018983999888102215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,2048,0.016754666964213055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,1536,0.013381332986884646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,1536,0.01517955627706316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,1536,0.014363555444611443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,1024,0.009918221996890174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,1024,0.010973333484596677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,1024,0.011968889170222811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,768,0.009313777916961247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,768,0.009129777550697327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,768,0.010848888920413123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,512,0.007820444802443186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,512,0.007078222102589077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,512,0.009675555759006077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,256,0.005085333353943295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,256,0.007220444579919179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,128,0.0046897778908411665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,128,0.003456888927353753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,128,0.006873777757088344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,64,0.0031600000543726813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,64,0.006877333339717653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,16384,32,0.003175111073586676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,16384,32,0.006826666494210561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,16384,0.08224355512195163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,16384,256,0.0053084443012873335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,65536,0.42318399747212726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,12288,0.07945866717232598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,65536,0.28304889467027455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,65536,0.25043733914693195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,16384,0.10599555571873982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,12288,0.06165066692564222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,16384,0.06331288814544678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,12288,0.047298666503694325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,10240,0.0671351088417901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,10240,0.05377511183420817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,10240,0.04221777783499824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,8192,0.042824000120162964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,8192,0.05363555418120491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,8192,0.034373333056767784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,7168,0.03959822323587205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,7168,0.04807733496030172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,7168,0.03028800090154012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,6144,0.032885332902272545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,6144,0.0432933337158627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,6144,0.02793688906563653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,5120,0.027792887555228338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,5120,0.035445332527160645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,5120,0.024698666400379602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,4096,0.024296000599861145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,4096,0.028044445647133723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,4096,0.020374221934212577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,3584,0.021893334057595994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,3584,0.025183111429214478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,3584,0.018931556079122756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,3072,0.01938488913906945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,3072,0.02221244408024682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,3072,0.01699555582470364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,2560,0.01834844383928511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,2560,0.018978666928079393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,2560,0.015879111157523263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,2048,0.013912889692518445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,2048,0.016010666886965435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,2048,0.013591110706329346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,1536,0.012821333275900947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,1536,0.011936888926559024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,1024,0.009282666775915358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,1024,0.009384888741705153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,1024,0.01032977799574534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,768,0.008661333057615492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,768,0.007885333564546373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,768,0.009533333281675974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,1536,0.013056889176368713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,512,0.006163555714819167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,512,0.006284444282452266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,512,0.007320000065697565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,256,0.005105777747101254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,256,0.003806222230195999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,256,0.006552000012662675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,12288,128,0.004655999855862724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,128,0.003321777822242843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,128,0.006229333165619109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,64,0.002987555538614591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,64,0.006143110907740063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,12288,32,0.003034666594531801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,12288,32,0.006169777777459886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,16384,0.09058844380908543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,16384,0.06307733058929443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,65536,0.37492977248297793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,16384,0.055192000336117215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,65536,0.2148195505142212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,65536,0.21897244453430176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,12288,0.06891822152667575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,12288,0.04353600078158908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,10240,0.05863110886679756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,10240,0.0396453340848287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,10240,0.03720266620318095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,8192,0.03171822097566392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,8192,0.04803644286261665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,8192,0.03073422114054362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,7168,0.028928889168633357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,7168,0.04272355635960897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,6144,0.024608888559871252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,7168,0.027822222974565294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,6144,0.03555377655559116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,6144,0.024969778127140466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,5120,0.032680888970692955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,12288,0.04570133487383524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,5120,0.0221760008070204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,4096,0.017779555585649278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,4096,0.027274666561020747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,4096,0.018568888306617737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,3584,0.018799111247062683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,3584,0.020962667134073045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,3584,0.017295110556814406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,3072,0.014893333117167154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,3072,0.018453333112928603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,3072,0.015944888194402058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,2560,0.01555288831392924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,2560,0.015912888778580558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,2560,0.01456622282663981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,2048,0.011723555624485016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,5120,0.02053955528471205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,2048,0.013215999636385175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,2048,0.012903999951150684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,1536,0.012334222594896952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,1536,0.010448888772063784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,1536,0.011404444773991903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,1024,0.00868888861603207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,1024,0.008232889076073965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,1024,0.010061333576838175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,768,0.008185777399275038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,768,0.0070462218589252895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,768,0.008908444808589088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,512,0.006056889063782162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,512,0.005731555736727185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,512,0.007060444189442529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,256,0.005215999980767568
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,256,0.003563555578390757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,256,0.006501333167155583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,10240,128,0.0047066667013698155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,128,0.0032435556252797446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,128,0.006159111029571957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,64,0.002921777880854077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,64,0.006117333140638139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,10240,32,0.0029315555261241067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,10240,32,0.006102222121424145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,65536,0.29916177855597603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,16384,0.05084177851676941
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,16384,0.0743137796719869
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,16384,0.05234400100178189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,12288,0.03961866762903001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,12288,0.05758755736880832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,65536,0.18519377708435059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,65536,0.18212532997131348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,12288,0.04279377725389269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,10240,0.03246400091383193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,10240,0.04844089017974006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,10240,0.03601244423124526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,8192,0.02718488872051239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,8192,0.03946933481428359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,8192,0.030543999539481267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,7168,0.024099555280473497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,7168,0.030516445636749268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,7168,0.027020444472630818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,6144,0.020967110991477966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,6144,0.026758222116364375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,6144,0.02384088933467865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,5120,0.018512000640233357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,5120,0.022873777482244704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,5120,0.021345777644051447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,4096,0.01610311037964291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,4096,0.01921422282854716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,4096,0.017883555756674874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,3584,0.01757777730623881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,3584,0.017085333665211994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,3584,0.016771554946899414
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,3072,0.013645333548386892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,3072,0.01519733336236742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,3072,0.015117333994971381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,2560,0.014687110980351767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,2560,0.013249778085284762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,2560,0.013688888814714221
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,2048,0.010789333118332757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,2048,0.011094222466150919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,2048,0.012278222375445895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,1536,0.011472889118724398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,1536,0.00925244473748737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,1536,0.011071999867757162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,1024,0.008019555773999956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,1024,0.007228444847795699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,1024,0.009292444421185387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,768,0.0070977773931291364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,768,0.006303999986913469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,768,0.0075866662793689305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,512,0.005715555614895291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,512,0.005039111193683413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,512,0.0069102222720781965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,256,0.004855999930037392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,256,0.0035084444615576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,32,0.0029804443733559716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,256,0.006445333361625671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,8192,128,0.004488000025351842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,128,0.0032151111712058387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,128,0.00609777785009808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,8192,64,0.002941333378354708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,64,0.00600622221827507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,65536,0.24750312169392905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,8192,32,0.006049777898523543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,16384,0.05272355675697327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,16384,0.06039110819498698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,16384,0.03920177618662516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,65536,0.1680426730049981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,12288,0.03672444489267137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,65536,0.1407582230038113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,12288,0.04672977659437391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,12288,0.03147466646300422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,10240,0.03183199961980184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,10240,0.04011822077963088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,10240,0.02737244466940562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,8192,0.02606577840116289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,8192,0.031114667654037476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,8192,0.024846222665574815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,7168,0.023002665903833177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,7168,0.027787556250890095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,7168,0.021180444293551978
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,6144,0.020785777105225455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,6144,0.024496000674035814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,6144,0.019353777170181274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,5120,0.01802399920092689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,5120,0.021431111627154883
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,5120,0.017312000195185345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,4096,0.015671999918089975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,4096,0.017830222845077515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,4096,0.014767110347747803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,3584,0.01587377819750044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,3584,0.015828443898095023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,3584,0.013755555782053204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,3072,0.013071111506885953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,3072,0.014127110441525778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,3072,0.012794666820102267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,2560,0.01386044422785441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,2560,0.012289777398109436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,2560,0.011929777761300405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,2048,0.010035555395815108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,2048,0.010517333116796283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,2048,0.010713777608341642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,1536,0.010781333678298526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,1536,0.008752889103359645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,1536,0.009777777724795872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,1024,0.006878222028414409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,1024,0.00685777763525645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,1024,0.007893333004580604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,768,0.006272888845867581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,768,0.005978666659858491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,768,0.006798222247097228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,512,0.005594666633341048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,512,0.004253333227501975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,512,0.00629688882165485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,256,0.004772444566090901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,256,0.0033644443998734155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,256,0.0058106668293476105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,7168,128,0.004319111092223061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,128,0.0031031111462248694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,128,0.00555377784702513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,64,0.00286311118139161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,64,0.005519111123349931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,7168,32,0.00286666676402092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,7168,32,0.005531555662552516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,65536,0.22384089893764922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,65536,0.16712444358401826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,65536,0.13412977589501276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,16384,0.05311200022697449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,16384,0.0384479992919498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,12288,0.03613866700066461
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,12288,0.04139822059207492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,12288,0.03049244483311971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,10240,0.03256266646915012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,10240,0.03418044580353631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,10240,0.026576888230111863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,8192,0.024872000018755596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,16384,0.050147556596332125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,8192,0.02808977829085456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,8192,0.022397332721286353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,7168,0.02425688836309645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,7168,0.025380444195535447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,7168,0.020062221421135795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,6144,0.02108000053299798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,6144,0.022654222117529974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,6144,0.018657777044508193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,5120,0.01718488832314809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,5120,0.018618666463428073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,5120,0.01644088824590047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,4096,0.015057777365048727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,4096,0.016584888100624084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,4096,0.014076444837782117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,3584,0.016659556163681876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,3584,0.014560000763999091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,3584,0.013403555585278405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,3072,0.012814222110642327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,3072,0.01297688898113039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,3072,0.012274666792816587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,2560,0.014293332894643148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,2560,0.011422221859296164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,2560,0.011378666592968835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,2048,0.010113777385817634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,2048,0.009866666462686326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,2048,0.010219555762079027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,1024,0.007038222418891058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,1536,0.008235555556085374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,1536,0.0092613332801395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,1024,0.00674311113026407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,1024,0.006484444356626934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,768,0.006598222172922558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,768,0.005643555687533484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,768,0.006618666566080517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,512,0.005695111221737332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,512,0.003867555409669876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,512,0.006194666855865055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,256,0.004955555415815777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,256,0.003310222178697586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,256,0.005704888867007361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,1536,0.011050666371981302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,128,0.003039999968475766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,128,0.005480888817045424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,64,0.0028124445428450904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,64,0.005440000030729506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,6144,32,0.0028400001012616684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,6144,32,0.005391110976537068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,65536,0.18564444118075898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,16384,0.046450667911105685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,65536,0.15341422292921278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,65536,0.10815644264221191
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,16384,0.04851644568973117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,16384,0.03180888957447476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,12288,0.037490665912628174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,12288,0.024988444315062627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,10240,0.027992000182469685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,10240,0.03214577833811442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,6144,128,0.00453955555955569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,10240,0.02218399941921234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,8192,0.024255999260478552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,8192,0.026791999737421673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,8192,0.018961777289708454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,7168,0.021239999267790053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,7168,0.024355555574099224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,7168,0.01715288890732659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,6144,0.020336889558368258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,6144,0.020750222934616935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,6144,0.01609155535697937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,5120,0.017088888419999015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,12288,0.0347360008292728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,5120,0.018279999494552612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,5120,0.014440889159838358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,4096,0.014070222775141398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,4096,0.015710221396552194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,4096,0.01257777793539895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,3584,0.014998222390810648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,3584,0.012103111379676394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,3584,0.011904888682895236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,3072,0.010947555303573608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,3072,0.013314666847387949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,3072,0.01110400011142095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,2560,0.01331999980741077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,2560,0.009731555150614845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,2560,0.01034755590889189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,2048,0.00977244476477305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,2048,0.00849511143234041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,2048,0.009501333038012186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,1536,0.010233778092596266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,1536,0.007259555160999298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,1536,0.008415111237102086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,1024,0.0063733334342638654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,1024,0.005967111223273807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,1024,0.006825777805513806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,768,0.006224888894293044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,768,0.005038222091065513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,768,0.006329777754015393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,512,0.005402666827042897
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,512,0.003693333516518275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,512,0.005982222242487802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,256,0.004610666798220741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,256,0.0032462223122517266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,256,0.005608888963858287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,5120,128,0.004287111262480418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,128,0.003003555453485913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,128,0.005426666802830166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,64,0.002744888886809349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,64,0.00543200017677413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,5120,32,0.0027582223216692605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,5120,32,0.005384888913896348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,65536,0.1494328843222724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,16384,0.04476622078153822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,65536,0.08588088883294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,16384,0.039439111948013306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,16384,0.02684177789423201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,12288,0.0313351121213701
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,12288,0.03086577852567037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,12288,0.02137511140770382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,10240,0.029283556673261855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,10240,0.026603556341595117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,7168,0.02015644477473365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,10240,0.01902577777703603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,8192,0.022809776994917128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,65536,0.15683110555013022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,8192,0.022231111923853557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,8192,0.01663733356528812
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,7168,0.022864888111750286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,7168,0.015228443675571017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,6144,0.019732443822754752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,6144,0.018153778380817838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,6144,0.014215999179416232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,5120,0.016372443901167977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,5120,0.016137777103318107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,5120,0.01291911138428582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,4096,0.013929777675204806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,4096,0.014017777310477363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,4096,0.01128088848458396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,3584,0.01733688844574822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,3584,0.01256622208489312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,3584,0.01072533345884747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,2048,0.00941422250535753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,3072,0.012034667034943899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,3072,0.011425777441925473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,3072,0.010124444133705562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,2560,0.013067555924256643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,2560,0.00850311087237464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,2560,0.009563555320103964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,2048,0.007432888779375289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,2048,0.008937777744399177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,1536,0.006576889091067844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,1536,0.006992888947327931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,1024,0.006311111152172089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,1024,0.005648889061477449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,1024,0.006265777680608962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,768,0.0063573333124319715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,768,0.004277333203289243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,768,0.006024888820118374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,512,0.005516444643338521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,512,0.0035911111368073356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,512,0.005746666755941179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,256,0.004729777574539185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,256,0.0031351111829280853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,256,0.005482666608360078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,1536,0.009734222458468543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,4096,128,0.004274666723277834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,128,0.0029146667155954572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,128,0.005312888986534542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,64,0.0027155555370781156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,64,0.005314666777849197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,4096,32,0.0027359999302360746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,4096,32,0.005233777893914117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,65536,0.12258577346801758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,65536,0.1444151136610243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,65536,0.07967822419272529
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,16384,0.03510311245918274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,16384,0.024675556355052527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,12288,0.03142044279310439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,12288,0.026535999443795946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,12288,0.021535999245113794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,10240,0.026628444592158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,10240,0.023072888453801472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,10240,0.01812977757718828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,8192,0.022825777530670166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,8192,0.02016088863213857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,16384,0.041435556279288396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,8192,0.015633778439627755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,7168,0.01972888906796773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,7168,0.017975111802419026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,7168,0.014512000812424554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,6144,0.018968888454967074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,6144,0.016208888755904306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,6144,0.013411555853154926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,5120,0.017921777235137094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,5120,0.014381332529915703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,5120,0.01238666640387641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,4096,0.013643555343151093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,4096,0.012895999683274163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,4096,0.010945777926180096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,3584,0.015818667080667283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,3584,0.01183377785815133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,3584,0.010611555642551847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,3072,0.010999110837777456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,3072,0.01146933353609509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,2560,0.008090666598743862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,3072,0.009863999982674917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,2560,0.012479110724396177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,2560,0.009404444032245213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,2048,0.008656889200210571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,2048,0.007177777588367462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,2048,0.008088889221350351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,1536,0.008919999831252629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,1536,0.0063546668324205614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,1536,0.006618666566080517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,1024,0.005930666708283954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,1024,0.004968889057636261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,1024,0.006232000059551663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,768,0.006009777800904379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,768,0.004212444441186057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,768,0.005969777703285217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,512,0.005325333525737126
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,512,0.003579555700222651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,512,0.005751111027267244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,256,0.004596444467703502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,256,0.0031511110977994073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,256,0.00545155546731419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3584,128,0.0041804446114434134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,128,0.0028568889117903183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,128,0.00527555536892679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,64,0.0026968889352348116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,64,0.0052897776994440295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3584,32,0.0027511111564106415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3584,32,0.005299555758635203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,65536,0.10686933332019383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,16384,0.042600890000661217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,16384,0.03074666526582506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,65536,0.1421626673804389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,65536,0.07545333438449435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,16384,0.023752000596788194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,12288,0.030862222115198772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,12288,0.024671110841963027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,12288,0.019377777973810833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,10240,0.02605777813328637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,10240,0.02182933357026842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,10240,0.01739377776781718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,8192,0.02188266648186578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,8192,0.018392889036072623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,8192,0.015049777097172208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,7168,0.02020711037847731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,7168,0.016720889343155753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,7168,0.014031999640994601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,6144,0.018744889232847426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,6144,0.015280889140235053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,6144,0.013099555340078143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,5120,0.016504888733228047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,5120,0.013861333330472311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,5120,0.012096888489193387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,4096,0.013328889177905189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,4096,0.012168000141779581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,4096,0.010676444404655032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,3584,0.011086222198274402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,3584,0.010188444621033138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,3072,0.012028444144460889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,3072,0.010284444524182213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,3072,0.009742221898502773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,2560,0.011737777955002256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,2560,0.007777777810891469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,2560,0.008886222210195329
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,2048,0.007896888587209914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,2048,0.006877333339717653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,2048,0.007191110816266801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,1536,0.00832711077398724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,1536,0.006152888966931238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,1536,0.006601777755551868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,1024,0.006262222097979651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,1024,0.004414222306675381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,1024,0.006151111175616582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,768,0.0058044443527857465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,768,0.003826666623353958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,768,0.005987555616431766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,3584,0.013253333667914072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,512,0.005150222116046482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,512,0.0035004446076022256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,512,0.0057297779454125305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,256,0.004483555754025777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,256,0.003142222141226133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,256,0.005469333380460739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,3072,128,0.0041306668685542215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,128,0.0028906667398081887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,128,0.005292444593376583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,64,0.0027102221631341507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,64,0.0052577778697013855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,16384,0.039643555879592896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,3072,32,0.002699555622206794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,3072,32,0.005246222019195557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,65536,0.09473777479595608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,65536,0.11737244658999973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,16384,0.026774222652117412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,65536,0.07120088736216228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,16384,0.02265600032276577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,12288,0.03104177779621548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,12288,0.021189333664046392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,12288,0.018515555395020377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,10240,0.02629599968592326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,10240,0.018445332845052082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,10240,0.016755556066830952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,8192,0.02036888897418976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,8192,0.01605866683853997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,8192,0.014579556054539151
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,7168,0.0184515549076928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,7168,0.014592889282438489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,7168,0.013584000368913015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,6144,0.01834844383928511
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,6144,0.01314399970902337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,6144,0.012736000120639801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,5120,0.015631111131774057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,5120,0.011915555430783166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,5120,0.011754666765530905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,4096,0.012237333589129977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,4096,0.010692444112565784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,4096,0.010279110736317104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,3584,0.012528888881206512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,3584,0.010072000324726105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,3584,0.010012444522645738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,3072,0.011115555961926779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,3072,0.009394666386975182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,3072,0.009166222479608323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,2560,0.010388444695207808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,2560,0.00740533322095871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,2560,0.007914666500356462
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,2048,0.007689778175618913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,2048,0.006679111056857639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,2048,0.0069431112044387394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,1536,0.007912889122962952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,1536,0.005591111050711737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,1536,0.006522666662931442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,1024,0.0058017778727743365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,1024,0.00453333349691497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,1024,0.00611111107799742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,768,0.0037875556283526947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,768,0.005573333137565189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,768,0.005881777654091517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,512,0.004922666483455234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,512,0.003474666840500302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,512,0.005692444327804778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,256,0.004388444539573458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,256,0.0030862221287356485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,256,0.005408888889683618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2560,128,0.0040995557275083335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,128,0.0028124445428450904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,128,0.0052693333062860705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,64,0.0028284444577164124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,64,0.0052595556610160405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2560,32,0.0026942222482628296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2560,32,0.0052355556852287715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,65536,0.07377955648634169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,65536,0.13844977484809026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,16384,0.03819466630617777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,65536,0.06835911009046766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,16384,0.022647110952271357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,16384,0.021858667333920796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,12288,0.028744889630211726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,12288,0.018422222799725003
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,12288,0.018047110901938546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,10240,0.026680888401137456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,10240,0.016042666302786935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,10240,0.016184000505341422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,8192,0.02300888962215847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,8192,0.013993778162532382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,8192,0.01402933398882548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,7168,0.01868355605337355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,7168,0.013136000268989138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,7168,0.013198222551080914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,6144,0.018001778258217704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,6144,0.012073777616024017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,6144,0.012330667012267642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,5120,0.014901333385043673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,5120,0.010866666833559671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,5120,0.01146399974822998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,4096,0.012510221865442065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,4096,0.00980444418059455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,4096,0.010135999984211391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,3584,0.014055111342006259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,3584,0.00907733374171787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,3584,0.009142222503821055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,3072,0.009854222337404886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,3072,0.008616000413894653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,3072,0.008119111259778341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,2560,0.011921777493423887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,2560,0.00721955547730128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,2560,0.007229333122571309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,2048,0.007923555870850882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,2048,0.0064239998658498125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,2048,0.006743999818960826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,1536,0.009314667019579146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,1536,0.004729777574539185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,1536,0.0063591111037466265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,1024,0.005967111223273807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,1024,0.00407288885778851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,512,0.005566222386227713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,1024,0.006021333237489064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,768,0.006183111005359226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,768,0.0037084445357322693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,768,0.005756444401211209
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,512,0.00535999983549118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,512,0.0033582221302721235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,256,0.003042666655447748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,256,0.0053048887186580235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,128,0.00424888895617591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,128,0.0027884443600972495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,128,0.00517155561182234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,64,0.0026480000880029467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,64,0.005236444373925527
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,2048,32,0.0026551110463009942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,2048,32,0.005154666801293691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,65536,0.05532977647251553
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,65536,0.11692355738745795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,16384,0.0355004436439938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,16384,0.018188445104493033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,65536,0.06699733601676093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,16384,0.021563554803530376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,12288,0.027991112735536363
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,12288,0.015028445257080926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,12288,0.017666666044129264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,2048,256,0.004605333424276776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,10240,0.024863999750879075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,10240,0.013654222091039022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,10240,0.015915556086434256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,8192,0.02072533302836948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,8192,0.011736888852384357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,8192,0.01390044391155243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,7168,0.018726223044925265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,7168,0.010976889067225985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,7168,0.013080000049538083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,6144,0.010107555323176915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,6144,0.012162666353914471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,5120,0.015122666954994202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,5120,0.009471110999584198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,5120,0.011231111155615913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,4096,0.012406222522258759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,4096,0.008754666480753157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,4096,0.0088746663596895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,3584,0.011427555647161273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,3584,0.00851466672288047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,3584,0.008095111283991072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,3072,0.00927111092540953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,3072,0.008103110724025303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,3072,0.007691555553012424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,2560,0.010317333042621613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,2560,0.009120889008045197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,2560,0.007272000114123027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,2048,0.0074373334646224976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,2048,0.00738844441043006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,2048,0.006804444309737947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,1536,0.008157333566082848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,1536,0.0063777777055899305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,6144,0.01661777827474806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,1536,0.006405333264006509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,1024,0.005616888817813661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,1024,0.005227555417352253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,1024,0.006038222048017714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,768,0.004331555631425646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,768,0.005816000203291575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,512,0.005024888863166173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,512,0.003920000046491623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,512,0.005600888695981767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,256,0.004313777718279096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,256,0.0033093332830402586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,256,0.0053386667536364655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,128,0.004057777838574516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,128,0.0029973333908451926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,128,0.005223999834722943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,64,0.0028124445428450904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,64,0.005167110926575131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1536,32,0.0028640000770489373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1536,32,0.005225777626037598
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,65536,0.04032977753215366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,65536,0.0924346645673116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,65536,0.06617155339982775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1536,768,0.005575110928879843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,16384,0.03377422359254625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,16384,0.014483556151390076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,16384,0.021299555897712708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,12288,0.024772443705134924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,12288,0.012258666257063547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,12288,0.017472889688279893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,10240,0.021751999855041504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,10240,0.011349333657158745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,10240,0.015839111473825242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,8192,0.018028444714016385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,8192,0.010071111222108206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,8192,0.013527111046844058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,7168,0.016453334026866488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,7168,0.009293333523803288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,7168,0.012531555361217923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,6144,0.014165333575672574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,4096,0.007753777835104201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,6144,0.008485332959228093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,6144,0.010658666491508484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,5120,0.012537777423858643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,5120,0.00848711116446389
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,5120,0.009658666948477427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,4096,0.010692444112565784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,4096,0.008402666283978356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,3584,0.009660444325870937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,3584,0.0075137780772315125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,3584,0.008065777520338694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,3072,0.008440000315507254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,3072,0.0069679998689227635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,3072,0.007635555333561367
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,2560,0.008761777646011775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,2560,0.011215111447705163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,2560,0.007248889240953658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,2048,0.006676444576846228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,2048,0.008416000339719985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,2048,0.006753777878151999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,1536,0.007043555378913879
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,1536,0.009152000149091085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,1536,0.006372444331645966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,1024,0.005157333281305102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,1024,0.007858666280905405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,1024,0.005976888868543837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,768,0.0051893335249688895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,768,0.005411555369695027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,768,0.00574933323595259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,512,0.005277333160241445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,512,0.005511111269394557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,256,0.004252444538805219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,256,0.003988444391224119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,256,0.005281777845488654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,128,0.003371555565132035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,128,0.005164444446563721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,64,0.0030746666921509635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,64,0.005165333135260476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,1024,32,0.0030622221529483795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,1024,32,0.005121777868933148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,512,0.004736888739797804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,65536,0.08083911074532403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,65536,0.03024266825781928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,65536,0.06611822048823039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,16384,0.011864888999197217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,16384,0.021222222182485793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,12288,0.020119110743204754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,12288,0.010119111173682744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,12288,0.01742044422361586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,10240,0.017422222428851657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,1024,128,0.00398577791121271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,10240,0.009644444617960189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,10240,0.01554844445652432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,8192,0.008914666871229807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,8192,0.012560000022252401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,16384,0.025535111625989277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,7168,0.013487999637921652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,7168,0.008485332959228093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,7168,0.011179555621412067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,6144,0.01145600030819575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,6144,0.007894222107198503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,6144,0.010330666270520952
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,5120,0.009381333159075843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,5120,0.007871111234029135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,5120,0.009554666777451834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,4096,0.009126221968068017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,4096,0.006868444383144379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,4096,0.008422222402360704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,3584,0.008626666333940294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,3584,0.007275555696752336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,3584,0.008068444828192392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,8192,0.015885333220163982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,3072,0.006572444405820634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,3072,0.007640000018808577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,2560,0.007735999921957652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,2560,0.008300444318188561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,2560,0.007274666594134436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,2048,0.006100444330109491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,2048,0.0068195557428730865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,2048,0.006813333266311222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,1536,0.0064177778032090925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,1536,0.006116444451941385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,1536,0.006421333385838403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,768,0.005803555664088991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,1024,0.004902222090297275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,1024,0.005972444597217772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,768,0.004840888910823398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,3072,0.007326222128338284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,768,0.004274666723277834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,512,0.004608889006906086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,512,0.0038337777886125776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,512,0.005500444521506627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,256,0.004126222183307012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,256,0.00327822214199437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,256,0.0053671110007498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,1024,0.005208000126812193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,128,0.003020444470975134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,128,0.00519466648499171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,64,0.0027866665687825945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,64,0.00518488883972168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,768,32,0.0028239999794297745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,768,32,0.005177777674463061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,65536,0.06896444161732991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,65536,0.02276266614596049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,65536,0.0653911100493537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,16384,0.02033422225051456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,16384,0.009856888817416297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,16384,0.020997333857748244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,12288,0.015807110402319167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,12288,0.008805333740181392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,12288,0.016238222519556682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,10240,0.013741333451535968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,10240,0.008467555873923832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,10240,0.013711999687883588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,768,128,0.003930666794379552
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,8192,0.011622221933470832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,8192,0.008067555725574493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,8192,0.011886222494973077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,7168,0.010683555569913654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,7168,0.007306666837798224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,7168,0.01109155515829722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,6144,0.010128888818952773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,6144,0.006751110984219446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,6144,0.010226666927337646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,5120,0.008811555802822113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,5120,0.0069057775868309875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,5120,0.009472889204819998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,4096,0.007778666913509369
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,4096,0.006436444405052397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,4096,0.008426666259765625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,3584,0.007375111182530721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,3584,0.0068195557428730865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,3584,0.008022222254011366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,3072,0.00685777763525645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,3072,0.006490666833188798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,3072,0.00759733302725686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,2560,0.006154666758245892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,2560,0.007229333122571309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,2048,0.0058559998869895935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,2048,0.005517333332035277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,2048,0.006747555401590135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,1536,0.0058408888677755995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,1536,0.004818666726350784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,1536,0.006319111006127463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,1024,0.004921777794758479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,1024,0.0041582224269707995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,1024,0.005985777825117111
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,768,0.0046693334976832075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,768,0.0037946667936113146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,2560,0.006917333437336816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,768,0.00573600000805325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,512,0.00443111111720403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,512,0.0034844444857703317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,512,0.005515555540720622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,256,0.004035555654101902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,256,0.0031146665828095544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,256,0.005274666680230035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,512,128,0.0038417776425679526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,128,0.0028880000528362063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,128,0.005145777844720417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,64,0.0027182222240500977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,64,0.005156444592608346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,512,32,0.002732444347606765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,512,32,0.005149333427349727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,65536,0.06740622387992011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,65536,0.014702222413486905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,65536,0.06503555509779188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,16384,0.019133332702848647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,16384,0.007927999728255803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,16384,0.018798222144444782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,12288,0.015415999624464246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,12288,0.006729777902364731
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,12288,0.015167999598715039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,10240,0.013530666629473368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,10240,0.006609777609507243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,10240,0.013612444202105204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,8192,0.011576889289749993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,8192,0.006618666566080517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,8192,0.011759999725553723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,7168,0.010743111371994019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,7168,0.006445333361625671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,5120,0.009359999663299983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,7168,0.010954666468832227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,4096,0.00774133370982276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,6144,0.009668444593747457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,6144,0.006325333482689328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,6144,0.010125333236323463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,5120,0.008741333252853816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,5120,0.006618666566080517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,4096,0.006313777632183499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,4096,0.008278222547637092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,3584,0.007335999773608313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,3584,0.006763555523422029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,3584,0.007876444194051955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,3072,0.006770666688680649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,3072,0.0063573333124319715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,3072,0.00757422215408749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,2560,0.0063031112982167145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,2560,0.006068444500366847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,2560,0.007109333243634965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,2048,0.005757333503829108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,2048,0.005341333233647876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,2048,0.00664533343580034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,1536,0.005379555539952383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,1536,0.0046844445168972015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,1536,0.006256888724035687
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,1024,0.0047866664826869965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,1024,0.003995555556482739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,1024,0.005843555761708154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,768,0.004405333350102107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,768,0.003662222375472387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,768,0.005687110953860813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,512,0.004370666626426908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,512,0.00332622230052948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,512,0.00544888898730278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,256,0.003942222230964237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,256,0.002983111060327954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,256,0.005177777674463061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,256,128,0.003846222327815162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,128,0.002769777758253945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,128,0.005061333378156026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,64,0.002612444468670421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,64,0.005103111267089844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,256,32,0.002623999905255106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,256,32,0.005150222116046482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,65536,0.06688355737262301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,65536,0.011385777758227454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,65536,0.06315910816192627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,16384,0.01909955508179135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,16384,0.006341333190600078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,16384,0.018577777677112155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,12288,0.015466666883892484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,12288,0.006280000011126201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,12288,0.015125332607163323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,10240,0.013335111240545908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,10240,0.00647288892004225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,10240,0.013559111290507846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,8192,0.011526222030321756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,8192,0.00629688882165485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,8192,0.011703111231327057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,7168,0.010616888602574667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,7168,0.006177777631415262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,5120,0.009356444080670675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,7168,0.01088444474670622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,6144,0.009575111170609793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,6144,0.006098666538794835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,6144,0.010110222631030613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,5120,0.008644444247086843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,5120,0.006406222366624408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,4096,0.006024000131421619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,4096,0.007650666766696506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,4096,0.008250666989220513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,3584,0.007195555501514011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,3584,0.006424888968467712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,3584,0.007894222107198503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,3072,0.006704000135262807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,3072,0.00610133343272739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,3072,0.0075048887067370946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,2560,0.006204444501135085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,2560,0.005712000032265981
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,2560,0.007129777636792924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,2048,0.00573333352804184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,2048,0.005109333329730564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,2048,0.006626666833957036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,1536,0.005327111317051782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,1536,0.004456888884305954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,1536,0.006243555496136348
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,1024,0.004783111313978831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,1024,0.0037955554823080697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,1024,0.00583022211988767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,768,0.004371555729044808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,768,0.0034933334423436057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,768,0.005612444546487596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,512,0.0042453333735466
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,512,0.0033128888656695685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,512,0.005408000200986862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,256,0.003879111260175705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,256,0.0029893333299292457
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,256,0.00519466648499171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,4,128,128,0.0037368887828456033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,128,0.0027742222365405825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,128,0.005060444275538127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,64,0.002607999990383784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,64,0.005052444421582752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,128,32,0.002630222174856398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,128,32,0.005032000028424793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,65536,0.009208888643317753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,65536,0.061760001712375216
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,16384,0.006226666685607698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,16384,0.018453333112928603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,12288,0.006127999888526068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,12288,0.015051555302408008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,10240,0.006294222341643439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,10240,0.013541333377361298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,8192,0.006200888918505774
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,8192,0.011685333318180509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,7168,0.006073777874310811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,7168,0.010863111250930362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,6144,0.005991999887757831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,6144,0.010072000324726105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,3072,0.006154666758245892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,5120,0.00637155564294921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,5120,0.009303111169073317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,4096,0.005952888892756567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,4096,0.008249777886602614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,3584,0.006339555399285422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,3584,0.007858666280905405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,3072,0.007485333416197035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,2560,0.005621333503060871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,2560,0.007135111424658034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,2048,0.005018666800525454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,2048,0.006604444649484422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,1536,0.004395555704832077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,1536,0.006238222122192383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,1024,0.0037324445115195382
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,1024,0.0058017778727743365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,768,0.0034728890491856467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,768,0.00564533347884814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,512,0.003168888803985384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,512,0.005356444252861871
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,256,0.0029528888149393927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,256,0.005108444641033809
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,128,0.002765333279967308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,128,0.005105777747101254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,64,0.0026088888860411118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,64,0.005085333353943295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,64,32,0.0025946667624844448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,64,32,0.004984000076850255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,65536,0.008286221987671321
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,65536,0.05926577912436592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,16384,0.006133333262470033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,16384,0.018539556198649935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,12288,0.00618577789929178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,12288,0.01514933341079288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,10240,0.006301333506902059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,10240,0.013527111046844058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,8192,0.006175999840100606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,8192,0.011658666862381829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,7168,0.006074666563007567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,7168,0.01089777797460556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,6144,0.005919111271699269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,6144,0.010128888818952773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,5120,0.0063742221229606206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,5120,0.009333333207501305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,4096,0.005863111052248213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,4096,0.008209777375062307
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,3584,0.006441777778996362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,3584,0.007834666305118138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,3072,0.006002666635645761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,3072,0.007490666376219855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,2560,0.005596444424655702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,2560,0.007072889142566257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,2048,0.004997333304749595
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,2048,0.006584888945023219
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,1536,0.004420444369316101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,1536,0.006232000059551663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,1024,0.003765333443880081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,1024,0.0058373332851462895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,768,0.003439999909864532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,768,0.005631111148330901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,512,0.0031804444475306403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,512,0.005408000200986862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,256,0.0028364445186323593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,256,0.005160888863934411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,128,0.002624888800912433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,128,0.005048888838953442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,64,0.002554666664865282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,64,0.00500177757607566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,4,32,32,0.00249422217408816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,4,32,32,0.005039111193683413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,16384,0.30487110879686147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,16384,0.28161777390374076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,16384,0.5223289065890843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,12288,0.39175822999742294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,10240,0.3263182110256619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,12288,0.2162355581919352
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,10240,0.17183289262983534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,8192,0.281240012910631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,8192,0.14999466472201878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,12288,0.23390844133165148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,10240,0.19894756211174858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,8192,0.16198933124542236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,7168,0.23431287871466744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,7168,0.12863822778066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,6144,0.19692533546023896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,6144,0.10569066471523708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,7168,0.12734222412109375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,6144,0.11766221788194443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,5120,0.09898489051394993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,5120,0.16507644123501247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,5120,0.1042871144082811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,4096,0.07613955603705512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,4096,0.1330666674507989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,4096,0.08594489097595215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,3584,0.11569244331783718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,3584,0.07535288731257121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,2560,0.04995377858479818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,3584,0.07231999768151177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,3072,0.06038666433758206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,2560,0.05314133233494229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,3072,0.10095289018419053
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,3072,0.05913688739140829
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,2560,0.0837528904279073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,2048,0.040793776512145996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,2048,0.06812800301445855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,2048,0.0428693327638838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,1536,0.034854223330815635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,1536,0.052150222990247935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,1536,0.034028444025251604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,1024,0.025461332665549383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,1024,0.036425779263178505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,1024,0.025759999950726826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,768,0.019502222537994385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,768,0.027851555082533095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,768,0.02197955548763275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,512,0.015125332607163323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,512,0.019799111617936026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,512,0.018051554759343464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,256,0.011520889070298938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,256,0.011216888825098673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,256,0.01594044433699714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,65536,128,0.008626666333940294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,128,0.007401777638329401
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,128,0.01366222235891554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,64,0.005450666778617435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,64,0.011039999624093374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,65536,32,0.004313777718279096
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,65536,32,0.010934222075674268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,65536,0.28616711828443736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,16384,0.07638133234447904
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,65536,0.5665244526333278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,16384,0.08850489060084026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,65536,0.33634222878350156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,16384,0.137046217918396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,12288,0.058028446303473584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,12288,0.09466666645473903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,12288,0.06496800316704644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,10240,0.04878133204248217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,10240,0.09105244610044692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,10240,0.05687644746568468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,8192,0.06440622276730008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,8192,0.04569422205289205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,7168,0.03687377770741781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,7168,0.05693511168162028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,7168,0.041660444604025945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,6144,0.03131999903255039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,6144,0.04914933443069458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,6144,0.036680890454186335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,5120,0.028202666176689997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,5120,0.04162222146987915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,5120,0.031614220804638334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,4096,0.022035555707083807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,4096,0.034500443273120456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,8192,0.040375110175874494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,4096,0.026505778233210247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,3584,0.023620444867346022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,3584,0.030174222257402208
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,3584,0.023932443724738225
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,3072,0.020211555891566806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,3072,0.02639644510216183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,3072,0.021446221404605444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,2560,0.01665422154797448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,2560,0.02275733318593767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,2560,0.019305777218606737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,2048,0.014392000105645923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,2048,0.018798222144444782
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,2048,0.01686844395266639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,1536,0.013357333011097379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,1536,0.01661333276165856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,1536,0.0144195556640625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,1024,0.010977778169843884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,1024,0.01107022249036365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,1024,0.012096000214417776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,768,0.009190222455395592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,768,0.009211555951171452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,768,0.010750222537252637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,512,0.00777599960565567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,512,0.007159111400445302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,512,0.009507555928495195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,256,0.005220444252093633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,256,0.005160000175237656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,256,0.007175111108356052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,16384,128,0.0046968890560997855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,128,0.0035013332962989807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,128,0.006816000160243776
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,64,0.0031804444475306403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,64,0.0068195557428730865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,16384,32,0.003144888828198115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,16384,32,0.006785777707894643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,65536,0.28407732645670575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,16384,0.10708444648318821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,16384,0.07903289132648043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,65536,0.42107465532090926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,16384,0.06218844652175903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,12288,0.06265422370698717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,65536,0.2505013412899441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,10240,0.044693334235085376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,12288,0.08045688602659437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,12288,0.04739200075467428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,10240,0.06832622157202826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,10240,0.04959288901752896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,8192,0.04308977723121643
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,8192,0.05424000157250298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,7168,0.04039377636379666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,8192,0.0377475553088718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,7168,0.04792888959248861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,7168,0.03218844532966614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,6144,0.032878223392698504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,4096,0.023726221587922838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,6144,0.04148888919088576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,6144,0.027505778604083594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,5120,0.028601778878106013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,5120,0.03526399864090814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,5120,0.02421155571937561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,4096,0.028311110205120508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,4096,0.02026844521363576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,3584,0.024432000186708238
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,3584,0.025075554847717285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,3584,0.01887022289964888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,3072,0.01960355540116628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,3072,0.022114665971861944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,3072,0.016923555069499545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,2560,0.019096889429622226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,2560,0.015560888581805758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,2048,0.014512000812424554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,2048,0.015944000747468736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,2048,0.013797333670987023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,1536,0.013019555144839816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,1536,0.012816000315878125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,1536,0.012069332930776807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,1024,0.010478222535716163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,1024,0.009467555416954888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,1024,0.01035377797153261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,768,0.008638222184446123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,768,0.007896000312434303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,256,0.003875555677546395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,768,0.009486222432719337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,2560,0.015677332878112793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,512,0.006178666734033161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,512,0.00629688882165485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,512,0.007419555551475949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,256,0.005031110925806893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,256,0.006487111250559489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,12288,128,0.004515555583768421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,128,0.0033457777980301115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,128,0.00618488879667388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,64,0.0030497778207063675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,64,0.006174222048785951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,12288,32,0.002992000016901228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,12288,32,0.006158222340875202
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,65536,0.2049351135889689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,65536,0.37346932623121476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,16384,0.06010311179690891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,16384,0.09011377890904744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,12288,0.06844355662663777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,16384,0.055050667789247304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,12288,0.045588443676630654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,10240,0.05817777580685086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,12288,0.0427653325928582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,10240,0.03929955429501004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,65536,0.20694755183325875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,10240,0.03744355506367154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,8192,0.03207111027505662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,8192,0.04758844441837735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,8192,0.03094666533999973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,7168,0.02997866604063246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,7168,0.042464888758129544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,7168,0.02857066525353326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,6144,0.024813334147135418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,6144,0.037858668300840594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,6144,0.02529511186811659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,5120,0.022838221655951604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,5120,0.03144177794456482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,5120,0.022078222698635522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,4096,0.017656000124083627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,4096,0.026162667406929865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,4096,0.01889511115021176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,3584,0.0210826661851671
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,3584,0.02102222210831112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,3584,0.01716888944307963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,3072,0.018361777067184448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,3072,0.018481777773963082
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,3072,0.015795555379655626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,2560,0.013391999734772576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,2560,0.015876443849669564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,2560,0.014478221535682678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,2048,0.011689778003427716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,2048,0.013312000367376538
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,2048,0.012830221818553077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,1536,0.012082666986518435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,1536,0.010753778119881948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,1536,0.011617778076065911
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,1024,0.010099555883142684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,1024,0.008221333225568136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,1024,0.010085333552625444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,768,0.008485332959228093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,768,0.0070053330726093715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,768,0.009083555804358589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,512,0.006007110906971826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,512,0.005671999934646819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,512,0.007159111400445302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,256,0.005127111242877113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,256,0.003600888782077365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,256,0.0064791109826829694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,10240,128,0.004675555560323927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,128,0.0032142222755485107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,128,0.006165333506133821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,64,0.002903999967707528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,64,0.006138666636413998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,10240,32,0.0029404444826973807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,10240,32,0.006103111224042044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,65536,0.2925822205013699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,65536,0.1848968929714627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,65536,0.1823813385433621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,16384,0.05516711208555433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,16384,0.07379555702209473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,16384,0.05089066757096184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,12288,0.040241777896881104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,12288,0.06271466943952772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,12288,0.0410168899430169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,10240,0.03403822249836392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,10240,0.049271110031339854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,10240,0.03418044580353631
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,8192,0.026965333355797663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,8192,0.039367109537124634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,8192,0.029654221402274236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,7168,0.030551999807357788
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,7168,0.027150221996837195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,6144,0.021213332811991375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,6144,0.026621333426899377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,6144,0.02420355545149909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,5120,0.023001778456899855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,5120,0.021319111188252766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,4096,0.01622488929165734
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,4096,0.019369777705934312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,4096,0.018242667118708294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,3584,0.01920977731545766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,3584,0.017072000437312655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,3584,0.016721778445773654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,3072,0.01752711170249515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,3072,0.015406222807036506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,3072,0.015190222197108798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,2560,0.0123831108212471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,2560,0.013264888690577613
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,2560,0.013774221969975365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,2048,0.011632888681358762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,2048,0.011236444115638733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,7168,0.02602044410175747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,2048,0.012111999922328524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,1536,0.011375111010339526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,1536,0.009211555951171452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,1536,0.01109688894616233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,1024,0.009451555709044138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,1024,0.0075146663520071245
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,5120,0.02100533412562476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,1024,0.00960355583164427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,768,0.006718222051858902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,768,0.006353777729802662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,768,0.007918222082985772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,512,0.00562755556570159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,512,0.005191111316283544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,512,0.0069822221994400024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,256,0.004790222065316306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,256,0.003540444291300244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,32,0.0029404444826973807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,256,0.0064479998416370815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,8192,128,0.004357333398527569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,128,0.003216000066863166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,128,0.006069333189063602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,8192,64,0.002938666691382726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,64,0.006052444378534953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,8192,32,0.006000888844331105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,65536,0.24272712071736655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,16384,0.05151822169621786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,65536,0.14888089232974583
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,16384,0.06004177861743503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,16384,0.04007644454638163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,12288,0.040363556808895536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,12288,0.04622311062282986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,12288,0.03158400124973721
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,10240,0.036836445331573486
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,10240,0.04135200050142076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,10240,0.027501333090994094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,8192,0.027407111393080816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,8192,0.031785776217778526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,8192,0.02337155573897892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,65536,0.17071733209821913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,7168,0.024104000793562994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,7168,0.028279112444983587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,7168,0.020997333857748244
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,6144,0.02068444424205356
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,6144,0.02495644489924113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,6144,0.019388443893856473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,5120,0.01923822197649214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,5120,0.021368000242445204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,5120,0.017245334055688646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,4096,0.01555999947918786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,4096,0.01771022213829888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,4096,0.014684443672498068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,3584,0.01886666648917728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,3584,0.015666666958067153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,3584,0.013929777675204806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,3072,0.016177778442700703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,3072,0.014173333843549093
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,3072,0.012816000315878125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,2560,0.011544889046086205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,2560,0.012448000411192576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,2560,0.011855111353927188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,2048,0.010063110954231685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,2048,0.010592888626787396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,2048,0.01074222226937612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,1536,0.010475555227862464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,1536,0.008868444297048781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,1536,0.009716444545321995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,1024,0.007988444632954067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,1024,0.0069751110341813825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,1024,0.00813066628244188
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,768,0.006257777826653586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,768,0.006019555446174409
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,768,0.006816888848940532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,512,0.005407111098368962
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,512,0.004390222330888112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,512,0.006254222244024277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,256,0.004655999855862724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,256,0.0034257777863078644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,256,0.0057653333577844836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,7168,128,0.004311999926964442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,128,0.003096888876623578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,128,0.005511111269394557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,64,0.002875555513633622
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,64,0.005511111269394557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,7168,32,0.002879111096262932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,7168,32,0.005449777675999536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,65536,0.22352621290418836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,65536,0.17472533384958902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,16384,0.04913333389494154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,16384,0.055495109823015004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,65536,0.13320799668629965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,16384,0.03870933254559835
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,12288,0.040672888358434044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,8192,0.026332444614834253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,12288,0.04068888889418708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,12288,0.0299928883711497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,10240,0.033835556772020124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,10240,0.026191110412279766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,8192,0.028869334194395278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,8192,0.021964444054497614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,7168,0.02795733345879449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,7168,0.025501334004931982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,7168,0.020192000601026747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,6144,0.019672888848516677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,6144,0.022264889544910852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,4096,0.016315556234783597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,6144,0.01850488947497474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,5120,0.02012622190846337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,5120,0.019825778073734708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,5120,0.016442666451136272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,4096,0.014917333920796713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,4096,0.014205333259370593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,3584,0.018452444010310702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,3584,0.014302222265137566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,3584,0.01330311099688212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,3072,0.015657777587572735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,3072,0.013111111190583972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,3072,0.012294222083356647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,2560,0.011865778101815118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,2560,0.011499555574523078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,10240,0.03304800060060289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,2560,0.011376000112957425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,2048,0.010176888770527309
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,2048,0.009958222508430481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,2048,0.010305778019958073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,1536,0.011027555498811932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,1536,0.008281778130266402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,1536,0.009422221945391761
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,1024,0.008047999607192146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,1024,0.006528000036875407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,1024,0.007050666544172499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,768,0.006451555424266391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,768,0.005669333454635408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,768,0.006627555522653792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,512,0.005555555638339784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,512,0.003920000046491623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,512,0.006140444427728653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,256,0.004848888764778773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,256,0.003335111050142182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,256,0.005724444571468566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,6144,128,0.004425777743260066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,128,0.003089777711364958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,128,0.005495999836259418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,64,0.0028568889117903183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,64,0.005408888889683618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,6144,32,0.0028195555011431375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,6144,32,0.00544888898730278
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,65536,0.185571551322937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,16384,0.045444445477591626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,16384,0.047896001074049205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,65536,0.10591999689737956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,16384,0.03159200151761373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,12288,0.03886399997605218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,12288,0.037295109695858426
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,12288,0.025599110457632277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,10240,0.030829333596759375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,10240,0.03179999854829576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,10240,0.022064889470736187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,8192,0.024858666790856257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,8192,0.02667733364635044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,8192,0.01908888916174571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,65536,0.15101422203911677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,7168,0.02408711115519206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,7168,0.01719999975628323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,6144,0.01977599991692437
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,6144,0.020703110429975722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,6144,0.01613066593805949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,5120,0.017695110705163743
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,5120,0.019512888458040025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,5120,0.01421866648726993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,4096,0.015581332974963717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,4096,0.014462222655614218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,4096,0.01258488910065757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,3584,0.017811555001470778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,3584,0.012190221912331052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,3584,0.01187999960449007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,7168,0.022878222995334204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,3072,0.013920888304710388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,3072,0.010992888775136737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,3072,0.011137777732478248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,2560,0.012679110798570844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,2560,0.009692444569534725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,2560,0.010326222413116032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,2048,0.009711999860074785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,2048,0.008564444051848518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,2048,0.009532444179058075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,1536,0.009949333137936061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,1536,0.007351111206743453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,1536,0.008389333056079017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,1024,0.006040888941950268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,1024,0.006703111032644908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,768,0.006125333408514659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,768,0.00489777781897121
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,768,0.006351999938488007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,512,0.0052915554907586836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,512,0.0038035557501845886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,512,0.005959110955397288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,256,0.004560889055331548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,256,0.0032435556252797446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,256,0.005595555735958948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,128,0.004301333179076512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,128,0.003018666679660479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,5120,1024,0.007423111134105259
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,128,0.005385777602593104
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,64,0.002756444530354606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,64,0.005430222385459476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,5120,32,0.0027706666539112725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,5120,32,0.005387555393907759
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,65536,0.14986666043599448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,65536,0.16626311673058405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,16384,0.04436622063318888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,16384,0.03931022352642483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,65536,0.0884560015466478
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,16384,0.026314665873845417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,12288,0.03294310967127482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,12288,0.03094044327735901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,12288,0.02146400014559428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,10240,0.02775466607676612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,10240,0.026185777452256944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,10240,0.01906222270594703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,8192,0.023936000135209825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,8192,0.02212711175282796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,8192,0.016350222958458793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,7168,0.02012355625629425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,7168,0.015136889285511441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,6144,0.019667555888493855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,6144,0.01814844376511044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,6144,0.013967111706733704
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,5120,0.01603822244538201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,5120,0.012783110969596438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,4096,0.013759999639458127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,4096,0.013750221994188098
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,4096,0.01126844435930252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,3584,0.01870933340655433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,3584,0.01240799989965227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,7168,0.021682666407691106
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,5120,0.018779555956522625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,3584,0.010732444624106089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,3072,0.014607111612955729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,3072,0.011380444798204633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,3072,0.01013777818944719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,2560,0.012803555362754397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,2560,0.008661333057615492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,2560,0.00959288908375634
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,2048,0.009434666898515489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,2048,0.007681777907742395
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,2048,0.008806222014957005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,1536,0.009370666411187913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,1536,0.006593777901596493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,1536,0.0068853331936730286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,1024,0.005648889061477449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,1024,0.006280000011126201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,768,0.006296000132958095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,768,0.004054222255945206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,768,0.005991999887757831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,512,0.0036106668412685394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,512,0.005741333381997214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,256,0.0046568889584806235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,1024,0.007772444850868649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,256,0.0031999999450312722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,256,0.005477333234416113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,512,0.0053395554423332214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,4096,128,0.004247111164861255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,128,0.0028862222615215513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,128,0.005281777845488654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,64,0.0027031112048361036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,64,0.005264889034960005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,4096,32,0.0027706666539112725
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,4096,32,0.0052693333062860705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,65536,0.1536346673965454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,65536,0.12373599741193984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,65536,0.07920533418655396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,16384,0.043402665191226535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,16384,0.03464355402522617
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,16384,0.024645333488782246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,12288,0.035495112339655556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,12288,0.026330666409598455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,12288,0.020021332634819877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,10240,0.025810667210155066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,10240,0.02308622168170081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,10240,0.018260444204012554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,8192,0.022355554832352534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,6144,0.01864177816443973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,8192,0.020118221640586853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,8192,0.015644444359673392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,7168,0.02389066749148899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,7168,0.01804266704453362
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,7168,0.01427288850148519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,6144,0.016303999556435477
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,6144,0.013413333230548434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,5120,0.01702844434314304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,5120,0.014583110809326172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,5120,0.012411555482281579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,4096,0.013620444469981723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,4096,0.01273333364062839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,4096,0.010829333629873065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,3584,0.016936888297398884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,3584,0.011911999848153857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,3584,0.0103511114915212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,3072,0.013644444445768992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,3072,0.01091466678513421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,3072,0.00980711148844825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,2560,0.011567110816637674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,2560,0.008185777399275038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,2560,0.009315555294354757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,2048,0.008847110801272923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,2048,0.007185777856243982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,2048,0.008080888953473832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,1536,0.00852177788813909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,1536,0.0063493334584765965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,1536,0.006981333096822103
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,1024,0.007125332951545715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,1024,0.004860444615284602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,1024,0.006194666855865055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,768,0.005903999838564131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,256,0.003141333245568805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,768,0.003934222377008862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,768,0.005963555640644497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,512,0.005191111316283544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,512,0.003536888708670934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,512,0.005752888818581899
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,256,0.004497777670621872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,256,0.005406222409672207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3584,128,0.004119111018048393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,128,0.0028444443725877335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,128,0.005286222116814719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,64,0.0027253333893087176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,64,0.0052391112678580815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3584,32,0.0027297778675953546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,65536,0.075436446401808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3584,32,0.00527022240890397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,65536,0.10850933525297378
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,65536,0.1341466638776991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,16384,0.03945422172546387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,16384,0.03143289022975498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,16384,0.023611555496851604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,12288,0.03372177812788222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,12288,0.024333332975705464
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,12288,0.019130667050679524
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,10240,0.026473777161704168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,10240,0.020766221814685397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,10240,0.017322666115230985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,8192,0.02202577723397149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,8192,0.018389332625601027
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,8192,0.015098666151364645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,7168,0.019973332683245342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,7168,0.01681511104106903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,7168,0.01400888959566752
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,6144,0.017976888351970248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,6144,0.015114666687117683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,6144,0.013119999733236102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,5120,0.01588355501492818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,5120,0.013544888959990608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,5120,0.012042666474978128
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,4096,0.012115555504957834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,4096,0.010501333408885531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,3584,0.015249777171346875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,3584,0.011195555329322815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,3584,0.010185778141021729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,3072,0.01036177741156684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,3072,0.00961599995692571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,2560,0.010980444649855295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,2560,0.00778311077091429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,2560,0.008985777695973715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,2048,0.007872000336647034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,2048,0.006917333437336816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,2048,0.0075084442893664045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,1536,0.008298666940795051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,4096,0.01293511109219657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,1536,0.006155555446942647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,1536,0.00654755574133661
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,3072,0.012679110798570844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,1024,0.006879111131032308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,1024,0.004363555461168289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,1024,0.006173333360089197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,768,0.0057848890622456866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,768,0.0038515557017591265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,768,0.00598933340774642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,512,0.005056000004212062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,512,0.003543111185232798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,512,0.005740444279379315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,256,0.00444177786509196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,256,0.003141333245568805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,256,0.005520888914664586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,3072,128,0.004007110993067424
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,128,0.0028737777223189673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,128,0.0052782222628593445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,64,0.0026933333526055017
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,64,0.0052328887912962176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,3072,32,0.0026942222482628296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,3072,32,0.005231110999981563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,65536,0.09360266394085354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,65536,0.12363733185662164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,65536,0.07083377573225233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,16384,0.026576888230111863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,16384,0.022450667288568284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,12288,0.03315111001332601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,12288,0.021288888321982488
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,12288,0.01847022275129954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,10240,0.027476444840431213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,10240,0.018616888258192275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,10240,0.01667644414636824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,8192,0.022456889351209004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,8192,0.015834665960735746
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,8192,0.0144177774588267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,7168,0.018197332819302876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,7168,0.014560000763999091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,16384,0.0388008885913425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,7168,0.013495999905798169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,6144,0.017161778277821012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,6144,0.013047110703256396
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,6144,0.0125928885406918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,5120,0.014664888381958008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,5120,0.011983111500740051
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,5120,0.011704888608720569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,4096,0.01334755536582735
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,4096,0.010828444527255164
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,4096,0.010276444256305695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,3584,0.01423377792040507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,3584,0.010076444182131026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,2560,0.007911110917727152
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,3584,0.009898666706350114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,3072,0.01111822244193819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,3072,0.009520000053776635
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,3072,0.008966222405433655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,2560,0.008787555827034844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,2560,0.007366221812036302
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,2048,0.006642666541867786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,2048,0.007784888976150089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,2048,0.006918222126033571
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,1536,0.00742133292886946
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,1536,0.00564533347884814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,1536,0.006513777706358168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,1024,0.006424888968467712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,1024,0.0041351111398802865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,1024,0.006063999815119638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,768,0.00552977787123786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,768,0.003814222084151374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,768,0.00590844452381134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,512,0.004859555512666702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,128,0.0028248888750871024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,512,0.0034551111360390982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,512,0.005626666463083691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,256,0.004278222305907143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,256,0.003080000066094928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,256,0.005422222117582957
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2560,128,0.003973333372010125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,128,0.0052533331844541765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,64,0.0026977778308921387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,64,0.005249777601824866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,65536,0.06819911135567559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2560,32,0.0027111110587914786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2560,32,0.005173333403136996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,65536,0.0748622218767802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,65536,0.14066221978929308
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,16384,0.03817333446608649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,16384,0.02254399988386366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,16384,0.021725333399242822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,12288,0.03420889046457078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,12288,0.01835911141501533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,12288,0.017859554953045316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,10240,0.02591466738118066
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,10240,0.016908443636364408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,10240,0.016213332613309223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,8192,0.022073777185546026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,8192,0.014194667339324951
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,8192,0.014012444350454541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,7168,0.024135111106766596
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,7168,0.013085333009560903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,7168,0.013117333253224691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,6144,0.016708445217874315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,6144,0.012055111428101858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,6144,0.012239110966523489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,5120,0.01724088854259915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,5120,0.01092977739042706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,5120,0.01146933353609509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,4096,0.01274577776590983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,4096,0.009878222313192155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,4096,0.010075555907355415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,3584,0.016484444340070088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,3584,0.009197333620654212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,3584,0.008995555341243744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,3072,0.014141332772043018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,3072,0.00869066682126787
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,3072,0.007840000092983246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,2560,0.010585777461528778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,2560,0.007224000162548489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,2560,0.007248889240953658
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,2048,0.00887822194231881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,2048,0.0063982220987478895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,2048,0.006748444504208035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,1536,0.0090506664580769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,1536,0.0047155556579430895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,1536,0.006375999914275275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,1024,0.00756533361143536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,1024,0.004052444464630551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,1024,0.005973333285914526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,768,0.006149333384301927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,768,0.0037208890749348533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,768,0.00574311117331187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,512,0.00526311124364535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,512,0.0033608888172441055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,512,0.005534222142563925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,256,0.004472889006137848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,256,0.0030542222989930045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,256,0.0053075556125905775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,2048,128,0.0041697778635554844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,128,0.0028008888992998335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,128,0.005199111170238919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,64,0.002612444468670421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,64,0.005153777698675792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,2048,32,0.0026791110220882627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,2048,32,0.005129777722888523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,65536,0.054915553993648954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,65536,0.1238408883412679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,65536,0.06645600001017253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,16384,0.03458222084575229
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,16384,0.01844800015290578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,16384,0.02144355575243632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,12288,0.02939288814862569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,12288,0.01513777838812934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,12288,0.017557332913080852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,10240,0.02464177707831065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,10240,0.013880000346236758
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,10240,0.01590133375591702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,8192,0.020629333125220407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,8192,0.011920000116030375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,8192,0.013779555757840475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,7168,0.01851466629240248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,7168,0.011110222174061669
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,7168,0.013008889224794177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,6144,0.01645155582163069
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,6144,0.010124444133705562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,6144,0.012145777543385824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,5120,0.014705777168273926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,5120,0.009415999882751042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,5120,0.011181332998805575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,4096,0.011164444188276926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,4096,0.00888888869020674
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,4096,0.008454221818182204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,3584,0.013330666555298699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,3584,0.008563555777072906
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,3584,0.008064889245563084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,3072,0.010595555934641095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,3072,0.00810933361450831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,3072,0.007652444144090016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,2560,0.009523555636405945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,2560,0.009299555586444007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,2560,0.007243555453088548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,2048,0.0075128889746136135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,2048,0.007819555699825287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,2048,0.006822222222884496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,1536,0.007807999849319458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,1536,0.006213333457708359
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,1536,0.006395555618736479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,1024,0.006579555571079254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,1024,0.005252444495757421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,1024,0.005960000058015187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,768,0.005600000007285013
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,768,0.004518222063779831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,768,0.0057848890622456866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,512,0.004865777575307422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,512,0.0038951109680864546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,512,0.005567111074924469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,256,0.004307555655638377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,256,0.0032702222880389956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,256,0.005324444423119227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1536,128,0.003963555726740095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,128,0.003000000077817175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,128,0.005193777796294955
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,64,0.0028364445186323593
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,64,0.005221333354711533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1536,32,0.0027591110103660156
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1536,32,0.005229333208666907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,65536,0.039317333035998873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,65536,0.09784444173177083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,65536,0.06609600120120578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,16384,0.030091553926467896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,16384,0.01463733282354143
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,16384,0.02113955550723606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,12288,0.012298666768603854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,12288,0.017456889152526855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,10240,0.011167111496130625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,10240,0.01568000018596649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,8192,0.018645332919226754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,8192,0.010071111222108206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,8192,0.013472889032628802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,7168,0.017380444539917838
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,7168,0.009282666775915358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,7168,0.011700444751315646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,6144,0.008403555386596257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,6144,0.01129066695769628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,5120,0.010966222319338055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,5120,0.008368000388145447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,12288,0.02657777733272976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,5120,0.009455111291673448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,10240,0.022625777456495497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,4096,0.010423999693658618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,4096,0.007923555870850882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,4096,0.008400888906584846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,3584,0.011579555769761404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,3584,0.007705777883529663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,3584,0.007972444097201029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,3072,0.0068959999415609576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,3072,0.007610666255156199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,2560,0.006843555718660355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,2560,0.013146667016877068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,2560,0.00720000018676122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,6144,0.012834666503800286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,2048,0.007166222565703922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,2048,0.011796444654464722
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,2048,0.006789333290523953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,1536,0.00925422211488088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,1536,0.006351111249791251
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,1024,0.007810666329330868
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,1024,0.005946666830115848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,768,0.005158222383923001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,768,0.006044444524579578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,768,0.005771555420425203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,3072,0.009135999613338048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,512,0.004593777573770947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,512,0.0052693333062860705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,512,0.005504000104135937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,256,0.004226666771703296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,128,0.005155555489990446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,256,0.004005333201752769
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,256,0.005311111195219888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,1536,0.006577777779764599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,128,0.003912000192536248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,128,0.0033235556135574975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,64,0.003079111170437601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,64,0.005138666679461797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,1024,32,0.003034666594531801
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,1024,1024,0.005807111246718301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,1024,32,0.005180444568395615
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,65536,0.030065778228971694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,65536,0.06571022007200453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,16384,0.025849777791235182
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,16384,0.011932444241311816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,16384,0.021144888467258875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,12288,0.021372444099850122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,12288,0.010133333504199982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,12288,0.017315554949972365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,10240,0.017864000466134813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,10240,0.009717333647939894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,10240,0.015473778049151102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,65536,0.08278133471806844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,8192,0.014671999547216626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,8192,0.008984888593355814
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,8192,0.012390221986505719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,7168,0.013012444807423485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,7168,0.008385777473449707
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,7168,0.01128888875246048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,4096,0.008757332960764566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,6144,0.01088711122671763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,6144,0.007918222082985772
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,6144,0.010284444524182213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,5120,0.00979377743270662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,5120,0.007814221911960179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,5120,0.009499555660618676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,4096,0.006784889019197888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,4096,0.008414222134484185
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,3584,0.009717333647939894
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,3584,0.0070062221752272705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,3584,0.008062221937709385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,3072,0.008032889001899296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,2048,0.006232888748248418
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,3072,0.006551110910044775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,3072,0.007623111208279927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,2560,0.007526222202512953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,2560,0.00832177781396442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,2560,0.007274666594134436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,2048,0.007711110843552484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,2048,0.006761777732107375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,1536,0.006281777802440856
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,1536,0.0058337777025169795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,1536,0.006411555740568373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,1024,0.0054551110499435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,1024,0.005057777795526716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,1024,0.005979555762476391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,768,0.004885333279768626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,768,0.004307555655638377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,768,0.005776888794369168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,512,0.004478222380081813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,512,0.0038008888562520346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,512,0.005544888890451855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,256,0.004036444342798657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,256,0.0032293332947625052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,256,0.005320000151793162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,768,128,0.003879111260175705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,128,0.0029804443733559716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,128,0.005128889034191768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,64,0.002830222249031067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,64,0.005166222237878376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,768,32,0.002761777697337998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,768,32,0.00519466648499171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,65536,0.06945333215925428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,65536,0.023123555713229712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,16384,0.020056888461112976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,65536,0.06590577628877428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,16384,0.009957333405812582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,16384,0.020891555481486853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,12288,0.016560888952679105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,12288,0.00869777798652649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,12288,0.01576266686121623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,10240,0.013817778064144982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,10240,0.008449777960777283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,10240,0.013728000223636627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,8192,0.011521778172916837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,8192,0.008061333662933773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,8192,0.011821333732869891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,7168,0.007264888948864407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,7168,0.011051555474599203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,6144,0.010124444133705562
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,6144,0.006685333533419504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,6144,0.010288000106811523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,5120,0.008640000389681922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,5120,0.006767999794748094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,5120,0.009408888717492422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,4096,0.008431110945012834
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,4096,0.006341333190600078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,4096,0.00835111074977451
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,3584,0.008562666674455007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,3584,0.006837333242098491
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,2560,0.006189333481921091
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,3584,0.007972444097201029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,3072,0.006956444846259222
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,3072,0.006447111152940326
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,3072,0.007576000359323289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,2560,0.006128888991143968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,2560,0.007174222005738153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,2048,0.006089777582221561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,2048,0.005493333356248007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,2048,0.006719111154476802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,1536,0.005699555493063397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,1536,0.0048062221871482
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,1536,0.006342222293217977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,7168,0.010656889114114972
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,1024,0.005170666509204441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,1024,0.004179555508825514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,1024,0.005868444426192178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,768,0.004655999855862724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,768,0.00378577783703804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,768,0.005659555395444234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,512,0.004289777742491828
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,512,0.0034942221310403612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,512,0.005468444277842839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,256,0.004039999925427967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,256,0.0031946665710873077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,256,0.0052622221410274506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,512,128,0.0038168889780839286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,128,0.0028497777465316984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,128,0.005157333281305102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,64,0.002716444432735443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,64,0.005184000151024925
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,512,32,0.00272444449365139
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,512,32,0.005110222018427319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,65536,0.014706666270891825
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,65536,0.06685955656899346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,65536,0.06551644537183973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,16384,0.019066666563351948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,16384,0.007868444754017724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,16384,0.01850488947497474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,12288,0.015464888678656684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,12288,0.006716444260544247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,12288,0.015192889504962497
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,10240,0.013355555633703867
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,10240,0.006628444625271692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,10240,0.013549333645237817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,8192,0.006559111177921295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,8192,0.011680888632933298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,7168,0.010602666272057427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,7168,0.006446222050322427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,7168,0.010919999745157031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,6144,0.00963466697269016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,6144,0.0063324446479479475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,6144,0.01015466699997584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,5120,0.008627555436558193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,5120,0.006716444260544247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,5120,0.009335999687512716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,4096,0.007672000262472365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,4096,0.0062871111763848206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,4096,0.008279110822412703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,3584,0.007273777491516537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,3584,0.00665511108107037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,3584,0.007928888830873702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,3072,0.006781333436568578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,3072,0.0064560001095136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,3072,0.007520889242490132
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,2560,0.00610133343272739
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,2560,0.006063111126422882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,2560,0.007106666763623555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,2048,0.005793777604897817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,2048,0.0053688887920644544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,2048,0.006647999915811751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,1536,0.00517155561182234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,1536,0.004670222186379963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,1536,0.006266666783226862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,1024,0.0047884442740016514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,1024,0.003999111139112049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,1024,0.005836444596449534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,768,0.00442844463719262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,8192,0.01149511088927587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,768,0.003653333418899112
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,768,0.005623111294375525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,512,0.0042160000238153665
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,512,0.003302222324742211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,512,0.005414222263627582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,256,0.003993777765168084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,256,0.002975111206372579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,256,0.005219555563396878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,256,128,0.0037724446091387006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,128,0.0027662221756246355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,128,0.005103111267089844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,64,0.002624888800912433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,64,0.005057777795526716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,256,32,0.0025840000145965153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,256,32,0.00507377791735861
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,65536,0.0665902230474684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,65536,0.011160888605647616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,65536,0.06300444735421075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,16384,0.019116444720162284
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,16384,0.006312888943486744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,16384,0.01850488947497474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,12288,0.015376889043384127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,12288,0.006132444573773279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,12288,0.0151582227812873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,10240,0.013402666482660504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,10240,0.006494222415818109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,10240,0.01354311075475481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,8192,0.011439999772442712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,6144,0.006017777654859755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,8192,0.006239111224810283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,8192,0.011711111499203576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,7168,0.010479999913109673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,7168,0.006131555471155379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,7168,0.01094044413831499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,6144,0.009600888523790572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,6144,0.010082667072614035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,5120,0.008645333349704742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,5120,0.006414222220579783
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,5120,0.009309333231714036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,4096,0.0076142218377855085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,4096,0.005983110931184556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,4096,0.008243555823961893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,3584,0.007087111473083496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,3584,0.006346666564544042
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,3584,0.007907555335097844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,3072,0.006703111032644908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,3072,0.006118222243256039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,3072,0.007477333148320516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,2560,0.006061333335108227
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,2560,0.005728000154097875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,2560,0.007099555598364935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,2048,0.005655111124118169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,2048,0.005110222018427319
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,2048,0.006629333313968446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,1536,0.005151999907361137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,1536,0.004471111214823193
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,1536,0.006271111054552927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,1024,0.0047457776963710785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,1024,0.003847111016511917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,1024,0.005843555761708154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,768,0.003481777591837777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,768,0.005600888695981767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,512,0.0041422223051389056
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,512,0.0033128888656695685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,512,0.0053386667536364655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,256,0.004121777911980947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,256,0.002971555623743269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,256,0.005151111218664381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,128,0.00371111101574368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,128,0.0027662221756246355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,128,0.005081777771313985
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,64,0.0026151111556424033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,64,0.005033777819739448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,128,32,0.0026488889836602737
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,2,128,768,0.004312889029582341
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,128,32,0.005021333280536863
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,65536,0.00889155517021815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,65536,0.06063466601901584
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,16384,0.006068444500366847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,16384,0.018530666828155518
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,12288,0.006121777825885349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,12288,0.015030221806632148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,10240,0.006258666515350342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,10240,0.01349066694577535
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,8192,0.006173333360089197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,8192,0.0116462219092581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,7168,0.006049777898523543
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,7168,0.010904889139864178
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,6144,0.005976888868543837
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,6144,0.010032888915803697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,5120,0.006305777778228124
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,5120,0.009292444421185387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,4096,0.005946666830115848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,4096,0.008207999997668797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,3584,0.0063991112013657885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,3584,0.007862221863534715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,1536,0.004401777767472797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,3072,0.006055111272467508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,3072,0.0074746666683091064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,2560,0.005597333527273602
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,2560,0.007052444749408298
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,2048,0.004970666848950916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,2048,0.006620444357395172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,1536,0.006218666831652324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,1024,0.003755555384688907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,1024,0.0058106668293476105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,768,0.0034755555291970572
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,768,0.005632888939645555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,512,0.003148444410827425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,512,0.00535377777285046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,256,0.00295733329322603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,256,0.005173333403136996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,128,0.0027546667390399506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,128,0.005028444445795483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,64,0.002607999990383784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,64,0.005037333402368757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,64,32,0.002605333303411802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,64,32,0.004996444616052839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,65536,0.008261333737108443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,65536,0.05956088834338718
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,16384,0.006112888869312074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,16384,0.018552000323931377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,12288,0.006067555397748947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,12288,0.015125332607163323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,10240,0.006233777850866318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,10240,0.01352533366945055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,8192,0.006175111151403851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,8192,0.011674666570292579
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,7168,0.006037333359320958
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,7168,0.010890666809346942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,6144,0.0058551111982928375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,6144,0.010096000300513374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,5120,0.006364444477690591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,5120,0.009324444664849175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,4096,0.005946666830115848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,4096,0.008250666989220513
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,3584,0.006342222293217977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,3584,0.007881777981917063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,3072,0.005998222364319696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,3072,0.007473777565691207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,2560,0.005625777774386936
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,2560,0.007091555330488417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,768,0.0034577778230110803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,2048,0.00498933345079422
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,2048,0.006607999818192587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,1536,0.004383999854326248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,1536,0.006237333433495627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,1024,0.0037617778612507712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,1024,0.0058151111006736755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,768,0.005587555468082428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,512,0.003148444410827425
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,512,0.005345777918895085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,256,0.0028168888141711554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,256,0.005168888717889786
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,128,0.0026275554878844153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,128,0.005049777941571342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,64,0.0025048889219760895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,64,0.004991111242108875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,2,32,32,0.002528888897763358
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,2,32,32,0.005050666630268097
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,16384,0.5206320020887587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,16384,0.28097955385843915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,16384,0.2831431229909261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,12288,0.21782399548424614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,12288,0.3914177682664659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,12288,0.22243555386861166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,10240,0.3261057800716824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,10240,0.17476267284817168
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,8192,0.263991117477417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,8192,0.13879556126064724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,10240,0.18772710694207084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,8192,0.15760266780853271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,7168,0.2401119867960612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,7168,0.13403022289276123
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,7168,0.12594578001234266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,6144,0.11535021993849014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,6144,0.19652888509962294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,5120,0.1647999948925442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,6144,0.11442311604817708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,5120,0.09405510955386692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,5120,0.09881866640514797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,4096,0.13338666492038304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,4096,0.07776888873842028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,3584,0.11574667029910618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,4096,0.08044533597098456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,3584,0.07495555612776014
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,3072,0.06549421946207683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,3584,0.07707555426491632
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,3072,0.1034311122364468
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,3072,0.06873599688212077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,2560,0.083661331070794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,2560,0.055065777566697865
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,2560,0.0582853356997172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,2048,0.04505688945452372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,2048,0.06769866413540311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,2048,0.05243555704752604
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,1536,0.03534400131967332
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,1536,0.05154044429461161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,1024,0.025751110580232408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,1536,0.04404000110096402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,1024,0.03604088889227973
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,1024,0.03678933448261685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,768,0.020784889658292133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,768,0.027518222729365032
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,768,0.03401599989997016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,512,0.015027556154463025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,512,0.01956977778010898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,512,0.03030844529469808
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,256,0.011439111497667102
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,256,0.026633777552180823
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,256,0.011251555548773872
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,65536,128,0.00865955568022198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,128,0.007453333172533248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,128,0.024754666619830664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,64,0.005647999958859549
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,64,0.020280889338917203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,65536,32,0.004320888883537716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,65536,32,0.020037333170572918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,16384,0.1368595494164361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,65536,0.5666746563381618
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,65536,0.2877768940395779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,16384,0.07769778039720324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,16384,0.07816977633370294
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,12288,0.09452266825570001
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,12288,0.05602844556172689
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,12288,0.0610026650958591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,65536,0.3051946692996555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,10240,0.05267822080188327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,10240,0.09066577752431233
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,10240,0.05272266599867079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,8192,0.06425778071085612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,8192,0.03992444607946608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,8192,0.043736000855763756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,7168,0.03762844536039565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,7168,0.06430577569537692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,7168,0.03965422179963853
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,6144,0.032010667853885226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,6144,0.04927555720011393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,6144,0.03837155633502536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,5120,0.02683111031850179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,5120,0.04163644380039639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,5120,0.031333333916134305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,4096,0.023240889112154644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,4096,0.03470666540993585
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,4096,0.027487110760476854
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,3584,0.02437866727511088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,3584,0.03028622269630432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,3584,0.025012445118692186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,3072,0.021405332618289526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,3072,0.02647022240691715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,3072,0.022908444205919903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,2560,0.019134221805466544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,2560,0.02274311085542043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,2560,0.021189333664046392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,2048,0.01608711149957445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,2048,0.01889600025282966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,2048,0.019144000278578866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,1536,0.013482666677898832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,1536,0.015073777900801765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,1536,0.016925333274735343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,1024,0.010787555740939246
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,1024,0.011066666907734342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,1024,0.014730667074521383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,768,0.009587555295891231
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,768,0.00907911111911138
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,768,0.013546666337384118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,512,0.0075084442893664045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,512,0.007108444141017065
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,512,0.012366222010718452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,256,0.005154666801293691
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,256,0.0047457776963710785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,256,0.00924800005224016
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,16384,128,0.004618666652176115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,128,0.003528000166018804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,128,0.008619555168681675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,64,0.0031635556370019913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,64,0.00865955568022198
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,16384,32,0.0031102223114834893
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,16384,32,0.008636444807052612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,65536,0.420801772011651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,65536,0.27822311719258624
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,16384,0.105204443136851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,16384,0.0800853305392795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,16384,0.060878223843044706
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,10240,0.06727021932601929
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,12288,0.08017066452238295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,12288,0.06200977828767565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,65536,0.24543645646837023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,12288,0.05183466606669956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,10240,0.05147822366820442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,10240,0.04229777720239428
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,8192,0.053469333383772105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,8192,0.034407999780442976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,7168,0.0401777790652381
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,7168,0.052574220630857676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,7168,0.03276177909639146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,6144,0.03375644485155741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,6144,0.041414221127827965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,6144,0.027540443672074214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,5120,0.030438221163219873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,5120,0.035102221700880266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,5120,0.024668445189793903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,8192,0.04310489032003614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,4096,0.023549333214759827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,4096,0.028549333413441975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,4096,0.02093244426780277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,3584,0.024332443873087566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,3584,0.025241777300834656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,3584,0.019362666540675692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,3072,0.02185955478085412
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,3072,0.0222053329149882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,3072,0.017997332745128207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,2560,0.018767999278174508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,2560,0.018936889039145578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,2560,0.016401777664820354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,2048,0.01646311084429423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,1024,0.010300444232092964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,2048,0.015966221690177917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,2048,0.014654222461912366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,1536,0.012844444149070315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,1536,0.012768000364303589
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,1536,0.013157332936922709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,1024,0.009496889180607265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,1024,0.011562666959232755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,768,0.00908088849650489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,768,0.007896000312434303
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,256,0.007377777662542131
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,768,0.010827555424637265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,512,0.006115555349323485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,512,0.006292444550328785
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,512,0.00812266684240765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,256,0.004982222285535601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,256,0.0038480001191298165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,12288,128,0.004580444345871608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,128,0.0033777778347333274
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,128,0.0069155556460221606
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,64,0.003043555551105075
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,64,0.006899555524190267
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,12288,32,0.00299822228650252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,12288,32,0.006904888898134232
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,16384,0.05988799863391452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,16384,0.08951199716991848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,65536,0.36669688754611546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,16384,0.05435644586881002
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,12288,0.04203111264440748
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,65536,0.21088622676001656
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,65536,0.2092684507369995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,12288,0.044216887818442456
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,12288,0.0685102211104499
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,10240,0.03952177696757846
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,10240,0.06435288985570271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,10240,0.036906666225857206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,7168,0.027971555789311726
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,8192,0.031889776388804116
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,8192,0.05274933245446947
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,8192,0.03277333246337043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,7168,0.0307004451751709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,7168,0.04213066564665901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,6144,0.02888266576661004
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,6144,0.03770666652255588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,6144,0.025242666403452556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,5120,0.02404622236887614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,5120,0.03253422180811564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,5120,0.022475555539131165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,4096,0.017453332742055256
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,4096,0.026802667313151892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,4096,0.019415999452273052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,3584,0.021687999367713928
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,3584,0.02083822256988949
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,3584,0.0181057784292433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,3072,0.01919644408755832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,3072,0.018360000517633226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,3072,0.016720889343155753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,2560,0.016365332735909354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,2560,0.01583377851380242
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,2560,0.015418666932317944
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,2048,0.014482667048772177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,2048,0.012901333471139273
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,2048,0.01382844481203291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,1536,0.011830222275522021
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,1536,0.010762666662534079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,1536,0.01260088880856832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,1024,0.010149333212110732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,1024,0.008199111455016667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,1024,0.01111644423670239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,768,0.008812444077597724
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,768,0.007048889166778988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,768,0.00983911090426975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,512,0.005912000106440649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,512,0.005715555614895291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,512,0.007846222155623967
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,256,0.004996444616052839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,256,0.003663111064169142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,32,0.0028968888024489083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,256,0.00721777809990777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,10240,128,0.004591111093759537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,128,0.00328177772462368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,128,0.006826666494210561
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,10240,64,0.0029475556479560006
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,64,0.006799110935793982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,10240,32,0.006783999916579988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,16384,0.07417333126068115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,65536,0.2923822138044569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,12288,0.05628444751103719
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,16384,0.04692444536421034
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,65536,0.20239644580417207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,65536,0.1693128877215915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,12288,0.037075555986828275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,12288,0.03767911261982388
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,10240,0.0357893341117435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,10240,0.04794222116470337
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,10240,0.03295822276009454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,8192,0.029675556553734675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,8192,0.0407955547173818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,8192,0.027987556325064764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,7168,0.027522666586769953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,7168,0.030645334058337744
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,7168,0.025417778227064345
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,6144,0.022712000542216833
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,6144,0.027260444230503503
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,6144,0.023028444912698533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,5120,0.018907555275493197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,5120,0.02311022248533037
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,5120,0.02069155540731218
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,4096,0.016037333342764113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,16384,0.05672888623343574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,4096,0.019335110982259113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,4096,0.017976888351970248
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,3584,0.02034133341577318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,3584,0.01705333259370592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,3584,0.01682222220632765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,3072,0.01813155578242408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,3072,0.01519733336236742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,3072,0.01573866605758667
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,2560,0.015508444772826301
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,2560,0.013378666506873237
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,2560,0.014818666709793938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,2048,0.013618666264745923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,2048,0.011557333171367645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,2048,0.013526221944226159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,1536,0.01129422254032559
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,1536,0.009329777624871995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,1536,0.011979555918110741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,1024,0.009178666604889764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,1024,0.007383999725182851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,1024,0.010453333457310995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,768,0.007127111156781514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,768,0.00629688882165485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,768,0.008361777497662438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,512,0.005520888914664586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,512,0.005190222213665645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,512,0.007507555186748505
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,128,0.006671111202902264
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,256,0.0047031111187405055
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,256,0.0034924443397257063
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,256,0.007012444237867991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,8192,128,0.004325333154863781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,128,0.0032275555034478507
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,64,0.0029493332323100832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,64,0.006701333241330252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,8192,32,0.0029813332690132987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,8192,32,0.006653333289755716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,65536,0.24441422356499565
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,16384,0.050620443291134305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,16384,0.05954133139716255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,65536,0.17255289024776885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,65536,0.14201688766479492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,16384,0.03679555654525757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,12288,0.03607022100024753
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,12288,0.04615555538071526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,12288,0.02906222144762675
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,10240,0.043175998661253184
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,10240,0.025455999705526564
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,8192,0.028134223487642076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,8192,0.03164533442921109
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,8192,0.021458667185571458
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,7168,0.02520177761713664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,6144,0.017542221479945712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,7168,0.02880177895228068
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,7168,0.019359111785888672
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,6144,0.02427644365363651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,6144,0.024847999215126038
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,5120,0.018748443987634447
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,5120,0.021264889174037512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,5120,0.015642666154437594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,4096,0.015797333584891427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,4096,0.017857778403494094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,4096,0.013554666605260638
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,3584,0.01863822175396813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,3584,0.015808888607554965
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,10240,0.03259733319282532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,3584,0.012541333006487953
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,3072,0.01680799987581041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,3072,0.014283556077215405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,3072,0.011893333660231696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,2560,0.01402933398882548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,2560,0.012396444049146442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,2560,0.010871111518806882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,2048,0.012301333248615265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,2048,0.01069777790043089
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,2048,0.009918221996890174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,1536,0.01034222212102678
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,1536,0.008785777621799046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,1536,0.008932444784376357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,1024,0.00794400026400884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,1024,0.006909333169460297
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,1024,0.0068764446510208984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,768,0.006643555644485686
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,768,0.006009777800904379
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,768,0.005960888746711943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,512,0.00546488869521353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,512,0.004418666578001446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,512,0.005547555370463266
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,256,0.00463733325401942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,256,0.0034142221427626083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,256,0.005222222043408289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,7168,128,0.004328000048796336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,128,0.003154666680428717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,128,0.005032888717121548
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,64,0.002862222285734283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,64,0.005025777965784073
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,7168,32,0.0028888889484935333
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,7168,32,0.004960000101062986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,65536,0.22325865427652994
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,65536,0.17180089155832926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,65536,0.1173422204123603
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,16384,0.05096088846524557
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,16384,0.05461511015892029
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,16384,0.033871110942628645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,12288,0.03561333484119839
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,12288,0.04257511099179586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,12288,0.02684177789423201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,10240,0.032680001523759626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,10240,0.03613777624236213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,10240,0.023744000328911677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,8192,0.027097778187857732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,8192,0.029525332980685767
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,8192,0.01973511113060845
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,7168,0.02593333356910282
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,7168,0.025010666913456384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,7168,0.0181377778450648
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,6144,0.02141600019401974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,6144,0.02274311085542043
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,6144,0.016156444946924847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,5120,0.017512889371977914
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,5120,0.01942488882276747
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,5120,0.01484444406297472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,4096,0.016509332590632968
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,4096,0.0127697777416971
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,3584,0.0195768889453676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,3584,0.01461866663561927
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,3584,0.012030222349696688
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,3072,0.012981333666377597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,3072,0.011159111228254108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,2560,0.011719110939237805
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,2560,0.010268444816271463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,2048,0.013175110850069257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,2048,0.010071111222108206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,2048,0.009370666411187913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,1536,0.010809777511490716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,1536,0.008208889100286696
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,1536,0.008607111043400234
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,1024,0.007756444315115611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,1024,0.006434666613737742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,768,0.007112888826264276
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,4096,0.017497777938842773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,1024,0.006253333141406377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,768,0.00563733321097162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,512,0.005526222288608551
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,768,0.005894222193294101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,512,0.005456888841258154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,512,0.004025777594910728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,3072,0.017504000001483493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,256,0.0046977777447965415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,256,0.0033528889632887314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,256,0.005179555465777715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,128,0.004399999976158142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,128,0.00309244439833694
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,6144,2560,0.01495733360449473
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,128,0.004980444494220945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,64,0.0028133332315418455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,64,0.004937777916590373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,6144,32,0.002850666642189026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,6144,32,0.004937777916590373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,65536,0.18989511330922446
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,16384,0.04677599999639723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,65536,0.16580711470709905
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,16384,0.04800088869200813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,65536,0.10247821940316094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,16384,0.029367112451129492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,12288,0.03459555572933621
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,12288,0.03691733214590285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,12288,0.02458488941192627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,10240,0.029179556502236262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,10240,0.03214755654335022
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,10240,0.021016000045670405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,8192,0.02456355591615041
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,8192,0.026335999369621277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,8192,0.018048889107174344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,7168,0.027297778262032404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,7168,0.02397777802414364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,7168,0.016435555285877652
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,6144,0.02053511142730713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,6144,0.020433778564135235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,6144,0.014929778046078153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,5120,0.01648533344268799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,5120,0.017993777990341187
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,5120,0.01352355546421475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,4096,0.014358222484588623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,4096,0.015557333827018738
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,4096,0.01179377817445331
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,3584,0.017840888765123155
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,3584,0.012392000191741519
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,3584,0.01110400011142095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,3072,0.015671110815472074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,3072,0.011022222538789114
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,3072,0.010409778190983666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,2560,0.013597333596812354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,2560,0.009759999811649323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,2560,0.009747555686367882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,2048,0.011933333343929715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,2048,0.008650666309727563
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,1024,0.006165333506133821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,2048,0.008994667066468133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,1536,0.007430222299363878
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,1536,0.009959110783206092
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,1536,0.007704888780911763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,1024,0.007256000406212277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,1024,0.00600000015563435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,768,0.006528888725572162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,768,0.005054222212897407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,768,0.0057555557125144536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,512,0.005254222287072076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,512,0.0037564444873068067
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,512,0.005428444594144821
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,256,0.004480888860093223
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,256,0.0032559999575217566
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,256,0.005156444592608346
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,32,0.004909333255555895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,5120,128,0.00425511101881663
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,128,0.002969777832428614
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,128,0.004980444494220945
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,64,0.002801777794957161
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,5120,64,0.004960000101062986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,5120,32,0.0027475555737813315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,65536,0.14642133977678087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,65536,0.153221329053243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,65536,0.0840666691462199
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,16384,0.03996977872318692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,16384,0.045935998360315956
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,16384,0.025116443634033203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,12288,0.030568887790044148
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,12288,0.030255999830034044
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,12288,0.020474667350451153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,10240,0.029863999949561224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,10240,0.02648799949222141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,10240,0.018412444326612685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,8192,0.02549244463443756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,8192,0.021930666433440313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,8192,0.015978667471143935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,7168,0.02390488816632165
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,7168,0.020261334048377145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,7168,0.014412444498803882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,6144,0.019708444674809773
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,6144,0.017994667092959087
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,6144,0.013265777793195514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,5120,0.020096888144810993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,5120,0.015993777248594496
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,5120,0.012064889073371887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,4096,0.013760888742076026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,3584,0.012553777959611682
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,4096,0.010623999767833285
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,4096,0.013790222505728403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,3584,0.018565333551830716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,3584,0.010135999984211391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,3072,0.016674665941132438
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,3072,0.011642666326628791
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,3072,0.009696000152164036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,2560,0.009027555584907532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,2048,0.012353777885437012
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,2560,0.00866844422287411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,2048,0.007937777373525832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,2048,0.008200888832410177
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,1536,0.006648889018429651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,1536,0.009380444056457942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,1536,0.0063662222690052455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,1024,0.0075599998235702515
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,1024,0.005635555419656966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,1024,0.005700444595681296
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,768,0.00702311098575592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,768,0.00408000002304713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,768,0.005423999908897612
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,512,0.005254222287072076
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,2560,0.014102222190962898
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,512,0.0036231109665499795
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,512,0.0052177777720822235
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,256,0.004567111117972268
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,256,0.0031786666562159858
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,256,0.005010666532648934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,128,0.002936888900068071
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,128,0.004829333474238713
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,64,0.002740444408522712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,64,0.004841777599520153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,4096,32,0.004821333206362194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,4096,32,0.002732444347606765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,65536,0.12231822808583577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,65536,0.1486497720082601
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,65536,0.07914933231141832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,16384,0.043613334496816
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,12288,0.025980444418059453
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,4096,128,0.004268444246715969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,16384,0.03506933318244086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,16384,0.024108444650967915
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,12288,0.032227555910746254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,10240,0.02697511182890998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,12288,0.01947377787695991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,10240,0.01740622189309862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,10240,0.022737777895397607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,8192,0.022504000200165644
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,8192,0.020067556036843192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,8192,0.015249777171346875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,7168,0.025847110483381484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,7168,0.017813333206706576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,7168,0.013935110635227628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,6144,0.018409777018758986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,6144,0.01629333363638984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,6144,0.012810666528013019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,5120,0.014469333820872836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,5120,0.011821333732869891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,4096,0.013928888572586907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,4096,0.012898666991127862
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,4096,0.010270222193664974
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,3584,0.016965332958433364
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,3584,0.011858666936556498
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,3584,0.009852444132169088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,3072,0.015615999698638916
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,3072,0.010822222464614444
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,5120,0.0162808895111084
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,3072,0.009301332963837517
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,2560,0.00814844419558843
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,2560,0.008845333423879411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,2048,0.011160000330872007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,2048,0.007314666277832455
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,2048,0.0074746666683091064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,1536,0.008595555192894405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,1536,0.00637511122557852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,2560,0.012963555753231049
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,1536,0.006161777923504512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,1024,0.007128888534175024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,1024,0.004777777940034866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,1024,0.005653333332803514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,768,0.006360000206364526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,768,0.003943110919660992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,768,0.005409777578380372
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,512,0.00508266646001074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,512,0.0035617777870761026
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,512,0.005126222140259213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,256,0.004461333155632019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,256,0.0031439999325407874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,256,0.004974222017659081
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3584,128,0.004090666770935059
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,128,0.0028737777223189673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,128,0.00481244424978892
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,64,0.0027137777457634607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,64,0.004782222211360931
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3584,32,0.00300799993177255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3584,32,0.004787555585304896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,65536,0.1074977782037523
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,65536,0.14307288328806558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,16384,0.041780445310804576
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,65536,0.07400088840060763
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,16384,0.030985779232449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,16384,0.022813333405388728
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,12288,0.030270222160551283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,12288,0.025028443998760645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,12288,0.01868977811601427
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,10240,0.021135111649831135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,10240,0.016601777738995023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,8192,0.022030222747060988
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,8192,0.018527110417683918
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,8192,0.01443199978934394
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,7168,0.021149333980348375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,7168,0.016346666547987197
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,7168,0.013335111240545908
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,6144,0.017978666557206046
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,6144,0.015135111080275642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,6144,0.012307555311255984
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,5120,0.015319999721315173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,5120,0.013665777941544851
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,5120,0.01145866678820716
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,4096,0.013239999612172445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,4096,0.012149333126015134
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,4096,0.009958222508430481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,3584,0.015761777758598328
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,10240,0.02493688960870107
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,3584,0.011120888921949597
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,3584,0.009558222360081142
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,3072,0.013815110756291283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,3072,0.010201777848932479
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,3072,0.009143110778596666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,2560,0.011683555940786997
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,2560,0.007744000189834171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,2560,0.008255110846625434
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,2048,0.00979377743270662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,2048,0.0068853331936730286
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,2048,0.006719111154476802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,1536,0.008089777496125963
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,1536,0.006125333408514659
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,1536,0.006102222121424145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,1024,0.006802666518423293
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,1024,0.004359111189842224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,1024,0.00562755556570159
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,768,0.006308444258239534
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,768,0.0038728887836138406
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,768,0.005410666680998272
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,512,0.004923555586073133
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,512,0.0035582222044467926
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,512,0.005159111072619756
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,256,0.004367111043797599
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,256,0.003170666595300039
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,256,0.004918222212129169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,3072,128,0.004054222255945206
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,128,0.002883555574549569
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,128,0.004779555731349521
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,64,0.0027137777457634607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,64,0.004785777793990241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,3072,32,0.0027457777824666765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,3072,32,0.0047662220895290375
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,65536,0.09393422471152411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,65536,0.11701599756876628
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,65536,0.07034755415386625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,16384,0.039439111948013306
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,16384,0.026705778307384912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,16384,0.021910222040282354
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,12288,0.028760890165964764
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,12288,0.021002666817771062
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,12288,0.017795556121402316
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,10240,0.02576622201336755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,10240,0.018237334158685472
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,10240,0.01606399979856279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,8192,0.023064000738991633
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,8192,0.01590577761332194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,8192,0.01385155568520228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,7168,0.02087822225358751
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,7168,0.014533332652515836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,7168,0.01295644458797243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,6144,0.019891555110613506
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,6144,0.013040000365840064
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,6144,0.012028444144460889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,5120,0.015984000431166757
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,5120,0.011930666863918304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,5120,0.01107022249036365
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,4096,0.012145777543385824
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,4096,0.010608000059922537
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,4096,0.00981955561372969
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,3584,0.014689778288205465
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,3584,0.010002666877375709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,3584,0.009490666290124258
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,3072,0.013109332985348172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,3072,0.009384888741705153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,3072,0.008839111361238692
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,2560,0.011076444553004371
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,2560,0.007391110890441471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,2560,0.007195555501514011
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,2048,0.0092622223827574
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,2048,0.00665244460105896
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,2048,0.006568888823191325
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,1536,0.007341333561473423
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,1536,0.005582222094138463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,1536,0.006072888771692912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,1024,0.0064080001579390625
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,1024,0.004209777961174647
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,1024,0.005653333332803514
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,768,0.0058959999846087555
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,768,0.0038319999972979226
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,768,0.005416000054942236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,512,0.004776000148720211
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,512,0.0035013332962989807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,512,0.005136000199450387
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,256,0.004319111092223061
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,256,0.0031155554784668815
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,256,0.004948444250557157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2560,128,0.003974222060706881
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,128,0.002814222127199173
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,128,0.0047208890318870544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,64,0.0026746667507621977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,64,0.004784000002675586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2560,32,0.002692444456948174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2560,32,0.004783111313978831
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,65536,0.14840000205569798
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,65536,0.07446044683456421
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,65536,0.06733422146903144
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,16384,0.040010667509502836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,16384,0.02254755629433526
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,16384,0.021048888564109802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,12288,0.027961777316199407
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,12288,0.018372444642914664
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,12288,0.017329777280489605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,10240,0.027600889404614765
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,10240,0.01646399994691213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,10240,0.015593777100245157
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,8192,0.02384177843729655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,8192,0.01423822177780999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,8192,0.01347644461525811
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,7168,0.02245866590076023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,7168,0.013402666482660504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,7168,0.012551111479600271
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,6144,0.0119991112086508
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,6144,0.011770666473441653
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,5120,0.014711111783981323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,5120,0.010961777634090848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,5120,0.010782221953074137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,4096,0.012474666866991254
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,4096,0.009848888549539778
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,4096,0.00920800036854214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,3584,0.01706044375896454
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,3584,0.009108444054921469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,3584,0.00869955536392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,6144,0.018346667289733887
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,3072,0.014727110664049784
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,3072,0.008661333057615492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,3072,0.007408888803588019
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,2560,0.007174222005738153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,2560,0.007092444433106317
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,2048,0.011149333582984077
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,2048,0.006452444526884291
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,2048,0.006554666492674086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,1536,0.0090844440791342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,1536,0.004808889081080754
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,1536,0.00611200018061532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,1024,0.0074035558435651995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,1024,0.004063111212518481
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,1024,0.00566133318675889
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,768,0.006718222051858902
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,768,0.0037395556767781577
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,2560,0.012305777933862476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,768,0.005379555539952383
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,512,0.005064888960785336
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,128,0.0027919999427265594
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,512,0.0033457777980301115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,512,0.005163555757866966
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,256,0.00443822228246265
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,256,0.0030453333424197305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,256,0.004949333353175057
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,2048,128,0.0041946665280395085
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,128,0.004777777940034866
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,64,0.0026657777941889236
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,64,0.004744889007674323
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,2048,32,0.0026702222724755607
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,2048,32,0.0047848886913723415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,65536,0.05459377831882901
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,65536,0.06439822249942355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,16384,0.03760088814629449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,16384,0.01834133267402649
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,16384,0.020389333367347717
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,12288,0.014885332849290637
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,12288,0.01682933337158627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,10240,0.02342222299840715
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,10240,0.013370667066839008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,10240,0.01516177753607432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,8192,0.0203813330994712
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,8192,0.011935999823941125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,8192,0.013175110850069257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,7168,0.021920889616012573
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,7168,0.010891555911964841
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,7168,0.012272000312805176
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,6144,0.01621955633163452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,6144,0.010100444157918295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,6144,0.011436444189813403
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,5120,0.014012444350454541
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,5120,0.009563555320103964
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,5120,0.010312000082598792
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,4096,0.011210666762457954
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,4096,0.008661333057615492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,4096,0.008068444828192392
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,3584,0.013882666826248169
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,3584,0.008585777547624376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,65536,0.11976710955301921
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,3584,0.007830222447713217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,3072,0.012249777714411417
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,3072,0.008088889221350351
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,3072,0.00739911115831799
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,2560,0.010520888699425591
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,2560,0.009445333646403419
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,2560,0.00700444479783376
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,2048,0.00942488925324546
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,2048,0.00813600007030699
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,1024,0.005229333208666907
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,2048,0.006510222123728857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,12288,0.027256000373098586
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,1536,0.007816000117195977
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,1536,0.006515555497672822
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,1536,0.006091555373536215
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,1024,0.006622222148709827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,1024,0.005640000104904175
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,768,0.005994666781690385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,768,0.004541333350870344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,768,0.005401777724424998
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,512,0.004774222357405556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,512,0.0038977778620190094
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,512,0.005132444616821078
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,256,0.004295111116435793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,256,0.0033111110743549135
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,256,0.004935111022657818
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1536,128,0.004017777740955353
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,128,0.003003555453485913
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,128,0.0048177776237328844
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,64,0.0028026666906144884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,64,0.0047600000268883174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1536,32,0.002817777709828483
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1536,32,0.004763555609517627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,65536,0.10687022076712714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,65536,0.04045866595374213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,65536,0.06235111422008938
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,16384,0.03185955683390299
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,16384,0.014359111587206522
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,16384,0.01982399986849891
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,12288,0.024262222978803847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,12288,0.01222222215599484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,12288,0.016338666280110676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,10240,0.02051466703414917
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,10240,0.011228444675604502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,10240,0.0147706667582194
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,8192,0.01961955593691932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,8192,0.009988444546858469
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,8192,0.012704888979593912
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,7168,0.018734221657117207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,7168,0.009207111265924241
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,7168,0.01163111130396525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,6144,0.015858666764365304
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,6144,0.008469333251317343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,6144,0.010187555518415239
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,5120,0.01148622234662374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,5120,0.008309333688682979
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,5120,0.009054222040706212
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,4096,0.010575999816258749
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,4096,0.007790221936172909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,4096,0.008072888685597314
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,3584,0.011361777782440186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,3584,0.007370666497283512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,2560,0.006970666348934174
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,3584,0.007781333393520779
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,3072,0.010536889235178629
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,3072,0.006823110911581252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,3072,0.007432888779375289
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,2560,0.008535111116038429
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,2560,0.014097778333557976
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,2048,0.008047999607192146
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,2048,0.012863111164834766
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,2048,0.0064604443808396654
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,1536,0.006875555548402999
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,1536,0.009746666583749983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,1536,0.006063111126422882
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,1024,0.005596444424655702
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,1024,0.007806222471925948
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,1024,0.005569777968857024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,768,0.005424889011515512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,768,0.006310222049554189
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,768,0.005375999957323074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,512,0.004505777938498391
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,512,0.0052737775776121355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,512,0.005126222140259213
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,256,0.0041644444896115195
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,256,0.003925333420435588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,256,0.00490133340160052
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,128,0.0033031110134389666
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,128,0.00480355570713679
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,64,0.003059555465976397
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,64,0.0047786666287316214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,1024,32,0.0030177777840031516
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,1024,32,0.0047360000511010485
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,65536,0.08148711257510714
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,65536,0.02998488810327318
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,16384,0.02499377727508545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,65536,0.06158755885230171
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,16384,0.011864888999197217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,16384,0.019504888190163504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,12288,0.020818667279349435
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,12288,0.010185778141021729
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,12288,0.016107555892732408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,10240,0.018152889278199937
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,10240,0.00959022260374493
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,10240,0.014136888914638095
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,8192,0.01519377695189582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,8192,0.008797333472304875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,8192,0.011306666665607028
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,7168,0.013183111117945777
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,7168,0.008372444245550368
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,7168,0.010623111493057676
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,6144,0.013997332917319404
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,1024,128,0.003915555361244413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,6144,0.00796622203456031
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,6144,0.009767110976907942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,5120,0.010151999692122141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,5120,0.007857778006129796
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,5120,0.009134222235944536
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,4096,0.008568888737095727
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,4096,0.006765333314736684
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,4096,0.008082666330867344
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,3584,0.010121777653694153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,3584,0.006855999843941794
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,3584,0.007755555212497711
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,3072,0.009012444151772393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,3072,0.006548444430033366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,3072,0.007373332977294922
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,2560,0.008030222521887885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,2560,0.008661333057615492
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,2560,0.007048000064161088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,2048,0.007639110916190677
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,2048,0.00648266656531228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,1536,0.006246222390068903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,1536,0.006278222219811545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,1536,0.006011555592219035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,1024,0.005044444567627377
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,1024,0.005550222264395819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,768,0.004373333520359463
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,768,0.005000000198682149
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,768,0.005365333209435145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,512,0.004457777573002709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,512,0.0038159998754660287
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,512,0.005112888912359874
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,256,0.004055110944641961
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,256,0.0032382222513357797
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,256,0.004906666775544484
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,128,0.003839111162556542
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,128,0.0029128889242808023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,128,0.004763555609517627
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,64,0.0028239999794297745
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,64,0.004774222357405556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,768,32,0.00281866660548581
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,768,32,0.0047208890318870544
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,65536,0.07235733005735609
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,65536,0.02261688808600108
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,16384,0.02013422217633989
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,65536,0.060517330964406334
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,16384,0.009825777676370408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,16384,0.018562666243977018
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,12288,0.01572533282968733
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,12288,0.008727110922336578
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,12288,0.015305777390797934
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,10240,0.01348533315791024
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,10240,0.008546666966544257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,10240,0.013620444469981723
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,2048,0.00741155528359943
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,768,1024,0.005459555735190709
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,8192,0.011510222322411008
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,8192,0.00794933322403166
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,8192,0.01126844435930252
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,7168,0.012574222352769641
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,7168,0.007048000064161088
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,7168,0.010491555763615502
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,6144,0.00965244405799442
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,6144,0.006565333240562015
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,6144,0.009696000152164036
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,5120,0.00846400029129452
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,5120,0.006731555693679386
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,5120,0.008982222113344405
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,4096,0.007753777835104201
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,4096,0.00629066675901413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,4096,0.007985778152942657
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,3584,0.008631111019187504
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,3584,0.006867555694447623
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,3584,0.007703999678293864
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,3072,0.007927999728255803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,3072,0.006554666492674086
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,3072,0.0073724447025193115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,2560,0.007070221834712558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,2560,0.006170666466156642
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,2560,0.0069057775868309875
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,2048,0.006607999818192587
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,2048,0.005523555394675996
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,2048,0.006405333264006509
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,1536,0.005517333332035277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,1536,0.004835555536879433
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,1536,0.005946666830115848
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,1024,0.005079111291302575
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,1024,0.0041680000722408295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,1024,0.005511999958091312
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,768,0.004774222357405556
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,768,0.0038151111867692736
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,768,0.0052684446175893145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,512,0.004293333325121137
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,512,0.0034702221552530923
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,512,0.004998222407367495
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,256,0.004049777984619141
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,256,0.003128888913326793
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,256,0.004769777672158347
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,512,128,0.003891555385457145
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,128,0.0028640000770489373
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,128,0.00464088883664873
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,64,0.002704888789190186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,64,0.004679999831649992
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,512,32,0.002691555561290847
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,512,32,0.0046942221621672315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,65536,0.06735910971959432
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,65536,0.01458577811717987
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,65536,0.060212446583641895
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,16384,0.01904444396495819
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,16384,0.007792889244026608
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,16384,0.01761777698993683
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,12288,0.015227556228637695
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,12288,0.006695111178689533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,12288,0.014409777190950183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,10240,0.013393777940008374
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,10240,0.00675022229552269
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,10240,0.012950222525331708
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,8192,0.011493333511882357
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,8192,0.006477333191368315
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,8192,0.011216888825098673
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,7168,0.010433778166770935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,7168,0.0064311110311084324
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,7168,0.010505778094132742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,6144,0.00960799968904919
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,6144,0.006329777754015393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,6144,0.009708444277445475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,5120,0.008466666771305932
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,5120,0.006600889066855113
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,5120,0.009073778159088558
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,4096,0.007731555236710443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,4096,0.00629066675901413
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,4096,0.008053333395057255
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,3584,0.007072889142566257
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,3584,0.006760889043410619
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,3584,0.007713778151406183
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,3072,0.006551110910044775
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,3072,0.00629333323902554
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,3072,0.007355555891990662
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,2560,0.006062222023804982
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,2560,0.006041777630647023
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,2560,0.0069351109365622205
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,2048,0.005684444473849402
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,2048,0.005375999957323074
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,2048,0.0063786668082078295
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,1536,0.005224888937340842
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,1536,0.004703999807437261
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,1536,0.005958222266700532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,1024,0.0046471113132105935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,1024,0.003996444245179494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,1024,0.005489777773618698
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,768,0.00442844463719262
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,768,0.003579555700222651
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,768,0.005277333160241445
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,512,0.0033191111352708605
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,512,0.005008888741334279
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,256,0.0039813332259655
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,256,0.0029599999801980127
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,256,0.0047848886913723415
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,128,0.0037911112109820047
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,128,0.0027671110712819626
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,128,0.0047191112405723995
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,64,0.0026142222599850762
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,64,0.004662222332424588
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,256,32,0.0026319999661710528
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,256,32,0.004693333473470476
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,65536,0.06623288657930163
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,65536,0.011400000088744693
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,65536,0.058244446913401283
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,16384,0.01901955571439531
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,16384,0.006242666807439592
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,16384,0.01753777762254079
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,12288,0.015218666858143277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,12288,0.006163555714819167
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,12288,0.014378666877746582
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,10240,0.013385777672131857
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,10240,0.006351999938488007
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,10240,0.01295644458797243
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,8192,0.011404444773991903
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,8192,0.006278222219811545
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,8192,0.011165333290894827
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,7168,0.010310221877362993
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,7168,0.006216889040337668
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,7168,0.010481778118345471
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,6144,0.009517333573765224
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,6144,0.006048000107208888
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,6144,0.009696889254781935
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,5120,0.008454221818182204
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,5120,0.0063991112013657885
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,5120,0.008974221845467886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,4096,0.0075866662793689305
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,4096,0.006000888844331105
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,4096,0.008016000191370646
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,3584,0.007038222418891058
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,3584,0.0063866666621632045
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,256,512,0.004115555435419083
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,3584,0.007661333514584436
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,3072,0.006535999890830781
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,3072,0.00611111107799742
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,3072,0.007285333342022366
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,2560,0.005664000080691443
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,2560,0.0069671107663048645
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,2048,0.005633777628342311
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,2048,0.005066666752099991
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,2048,0.0063973334100511335
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,1536,0.005107555538415909
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,1536,0.00444177786509196
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,1536,0.005884444548024072
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,1024,0.004584889031118817
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,1024,0.0037955554823080697
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,1024,0.005530666559934616
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,768,0.0034577778230110803
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,768,0.004397333496146732
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,768,0.005299555758635203
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,512,0.004104888687531154
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,512,0.003342222215400802
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,512,0.005009777843952179
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,256,0.003879111260175705
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,256,0.002973333415057924
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,256,0.0047377778424157035
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,128,0.0037413334680928122
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,128,0.002784888984428512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,128,0.004604444321658877
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,64,0.0026382222357723448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,64,0.0046533333758513136
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,128,32,0.0026151111556424033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,128,32,0.00490488898422983
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8_block,1,128,2560,0.006007110906971826
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,65536,0.008845333423879411
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,65536,0.05529955691761441
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,16384,0.006056889063782162
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,16384,0.017627555463049147
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,12288,0.006030222194062338
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,12288,0.014397333065668741
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,10240,0.006255110932721033
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,10240,0.012870222330093384
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,8192,0.006210666563775804
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,8192,0.011136888629860349
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,7168,0.006017777654859755
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,7168,0.010461333725187512
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,6144,0.005952000204059813
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,6144,0.009708444277445475
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,5120,0.0063368889192740125
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,5120,0.008973333570692275
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,4096,0.005958222266700532
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,4096,0.007981333467695449
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,3584,0.006337777607970768
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,3584,0.007630222373538547
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,3072,0.006078222145636876
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,3072,0.007335999773608313
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,2560,0.005592888842026393
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,2560,0.00683822234471639
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,2048,0.005013333426581489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,2048,0.00637511122557852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,1536,0.004420444369316101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,1536,0.005894222193294101
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,1024,0.003759110967318217
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,1024,0.005406222409672207
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,768,0.0034542222403817703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,768,0.005230222311284807
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,512,0.0031831111345026228
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,512,0.005033777819739448
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,256,0.002954666606254048
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,256,0.0047422221137417685
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,128,0.002749333365095986
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,128,0.004592888885074192
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,64,0.002609777781698439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,64,0.00463733325401942
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,64,32,0.00258577780591117
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,64,32,0.004632888982693355
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,65536,0.008302222523424361
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,65536,0.0557182232538859
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,16384,0.006092444476154115
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,16384,0.01761155492729611
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,12288,0.006116444451941385
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,12288,0.014354666074117025
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,10240,0.006258666515350342
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,10240,0.012902222573757172
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,8192,0.006088888893524806
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,8192,0.011309333145618439
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,7168,0.006075555665625467
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,7168,0.01050222251150343
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,6144,0.005950222412745158
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,6144,0.009740444521109263
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,5120,0.006320888797442119
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,5120,0.00903200027015474
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,4096,0.005959110955397288
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,4096,0.008073777788215214
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,3584,0.00637511122557852
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,3584,0.007672888537247975
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,3072,0.006081777728266186
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,3072,0.007335111498832703
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,2560,0.005577777822812398
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,2560,0.00693866651919153
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,2048,0.00499555551343494
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,2048,0.006406222366624408
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,1536,0.004396444393528832
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,1536,0.0059057776298787855
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,1024,0.0037475555307335327
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,1024,0.005432888865470886
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,768,0.0034773333205117118
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,768,0.0052284445199701525
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,512,0.00314666661951277
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,512,0.005013333426581489
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,256,0.0028026666906144884
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,32,0.00461422238085005
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,256,0.004770666774776247
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,128,0.0026311110705137253
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,128,0.004624888714816836
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,64,0.002581333327624533
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,fp8,1,32,64,0.004600888739029567
SGLang,0.5.6.post2,NVIDIA H200,gemm,sglang,float16,1,32,32,0.002567111204067866
