framework,version,device,op_name,kernel_source,mla_dtype,kv_cache_dtype,num_heads,batch_size,isl,tp_size,step,latency
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,1,1,0,0.5557719990611076
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,4,1,0,0.573243997991085
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,8,1,0,0.569568008184433
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,16,1,0,0.5722840055823326
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,32,1,0,0.5580319985747337
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,64,1,0,0.5661679953336716
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,128,1,0,0.5607000067830086
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,256,1,0,0.5748839974403381
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,512,1,0,0.5669959858059883
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,1024,1,0,0.5672880113124847
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,2048,1,0,0.6059960052371025
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,1,1,0,0.5712039992213249
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,4096,1,0,0.7015720009803772
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,4,1,0,0.5668960064649582
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,8192,1,0,1.1391559839248657
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,1,1,0,0.5584679916501045
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,8,1,0,0.5618479996919632
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,4,1,0,0.572320006787777
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,16,1,0,0.5570199936628342
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,8,1,0,0.5688319951295853
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,32,1,0,0.5642599985003471
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,16,1,0,0.562511995434761
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,64,1,0,0.5652960017323494
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,16384,1,0,2.632180005311966
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,32,1,0,0.5681560039520264
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,128,1,0,0.5568599924445152
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,64,1,0,0.5756720080971718
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,4,1,0,0.5811519995331764
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,256,1,0,0.561847984790802
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,128,1,0,0.5584319978952408
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,8,1,0,0.5860280022025108
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,512,1,0,0.5835840031504631
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,256,1,0,0.5747480019927025
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,16,1,0,0.5684200078248978
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,1024,1,0,0.6377600133419037
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,512,1,0,0.5686760097742081
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,32,1,0,0.5722160115838051
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,2048,1,0,0.7975639998912811
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,1024,1,0,0.5972039923071861
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,64,1,0,0.5677759945392609
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,2048,1,0,0.6738640069961548
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,4096,1,0,1.3246239721775055
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,128,1,0,0.5683399885892868
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,4096,1,0,0.9021119996905327
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,256,1,0,0.5781120136380196
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,512,1,0,0.5689279958605766
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,1024,1,0,0.5974640026688576
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,8192,1,0,1.765512004494667
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,2048,1,0,0.6389919966459274
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,8192,1,0,3.229908049106598
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,4096,1,0,1.005459986627102
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,8192,1,0,1.8359480053186417
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,16384,1,0,4.937620043754578
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,4,1,0,0.5727199986577034
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,8,1,0,0.5884439870715141
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,16,1,0,0.5783919915556908
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,32,1,0,0.5858560130000114
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,64,1,0,0.5837080031633377
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,16384,1,0,5.174920082092285
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,128,1,0,0.576324000954628
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,256,1,0,0.5807880088686943
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,4,1,0,0.5693200007081032
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,16384,1,0,10.366819977760315
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,512,1,0,0.6024680063128471
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,8,1,0,0.5687200054526329
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,1024,1,0,0.6443799957633018
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,4,1,0,0.5734960064291954
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,16,1,0,0.5642319992184639
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,8,1,0,0.5788640007376671
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,32,1,0,0.5710159987211227
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,2048,1,0,0.7936079949140549
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,16,1,0,0.5748640075325966
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,64,1,0,0.5665159970521927
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,4096,1,0,1.3527799993753433
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,32,1,0,0.5778720006346703
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,128,1,0,0.5783479958772659
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,256,1,0,0.5752799957990646
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,64,1,0,0.5662240087985992
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,128,1,0,0.5840279906988144
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,512,1,0,0.5936280116438866
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,256,1,0,0.5827639997005463
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,1024,1,0,0.6433280035853386
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,512,1,0,0.6407640054821968
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,8192,1,0,3.3300800025463104
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,2048,1,0,0.801864005625248
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,1024,1,0,0.7492880001664162
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,4096,1,0,1.4207040071487427
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,2048,1,0,1.113227978348732
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,4096,1,0,2.4175920486450195
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,8192,1,0,3.4931560158729553
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,16384,1,0,10.656983971595764
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,8192,1,0,6.472375988960266
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,4,1,0,0.633932001888752
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,8,1,0,0.587564006447792
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,16,1,0,0.5841759964823723
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,32,1,0,0.5738319978117943
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,64,1,0,0.5843279957771301
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,128,1,0,0.5842000097036362
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,256,1,0,0.6101760044693947
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,16384,1,0,10.694192051887512
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,512,1,0,0.6531400009989738
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,4,1,0,0.5679159983992577
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,1024,1,0,0.7473239973187447
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,8,1,0,0.5773439928889275
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,2048,1,0,1.1343719959259033
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,16,1,0,0.579804003238678
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,32,1,0,0.575887992978096
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,64,1,0,0.576044000685215
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,128,1,0,0.5827120095491409
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,4096,1,0,2.4583839178085327
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,256,1,0,0.6113719940185547
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,512,1,0,0.6322960034012794
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,1024,1,0,0.754816010594368
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,2048,1,0,1.2142640054225922
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,4096,1,0,2.6427760124206543
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,8192,1,0,6.66455602645874
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,16384,1,0,20.246576070785522
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,4,1,0,0.5807119980454445
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,8,1,0,0.5910639986395836
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,16,1,0,0.5861080065369606
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,32,1,0,0.57167599350214
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,64,1,0,0.5928599908947945
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,1,1,0,0.5745960026979446
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,128,1,0,0.6030760034918785
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,4,1,0,0.583392009139061
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,8192,1,0,7.126479983329773
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,256,1,0,0.6462680101394653
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,8,1,0,0.5694639906287193
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,512,1,0,0.7309359908103943
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,16,1,0,0.5649400129914284
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,32,1,0,0.5612919926643372
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,1024,1,0,1.0176440179347992
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,64,1,0,0.5748280137777328
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,128,1,0,0.5628360137343407
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,256,1,0,0.5738120004534721
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,2048,1,0,2.0049800127744675
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,512,1,0,0.5593319982290268
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,1024,1,0,0.5755840018391609
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,2048,1,0,0.5835840031504631
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,4096,1,0,0.6946319937705994
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,8192,1,0,1.1341239959001541
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,4096,1,0,4.623039901256561
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,16384,1,0,2.6259320080280304
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,4,1,0,0.572732001543045
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,16384,1,0,21.112435817718506
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,8,1,0,0.5834240093827248
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,16,1,0,0.5723559930920601
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,4,1,0,0.5881079882383347
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,32,1,0,0.5914560034871101
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,8,1,0,0.5990559980273247
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,64,1,0,0.5819360017776489
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,1,1,0,0.5563999935984612
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,16,1,0,0.5837960094213486
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,128,1,0,0.5832040086388588
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,4,1,0,0.5767680034041405
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,32,1,0,0.5912760123610497
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,256,1,0,0.5919280052185059
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,8,1,0,0.5569199919700623
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,64,1,0,0.5872199833393097
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,512,1,0,0.5901359841227531
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,16,1,0,0.5610399916768074
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,128,1,0,0.6168719977140427
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,1024,1,0,0.6109840124845505
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,32,1,0,0.5626279935240746
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,256,1,0,0.6534079983830452
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,2048,1,0,0.6434559971094131
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,64,1,0,0.5616319850087166
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,512,1,0,0.7198639959096909
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,128,1,0,0.550819993019104
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,4096,1,0,0.9299479946494102
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,256,1,0,0.5605920031666756
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,1,1,0,0.5673680081963539
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,1024,1,0,1.0393720120191574
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,16384,1,0,21.187983751296997
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,512,1,0,0.5596040114760399
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,4,1,0,0.5659360066056252
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,8192,1,0,12.981348037719727
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,1024,1,0,0.5878999903798103
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,8192,1,0,1.8539559990167618
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,8,1,0,0.5646080002188683
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,4,1,0,0.5941879972815514
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,2048,1,0,0.6424159929156303
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,2048,1,0,2.0349519848823547
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,16,1,0,0.5728239864110947
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,8,1,0,0.5980479940772057
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,32,1,0,0.5698520019650459
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,4096,1,0,0.8988119959831238
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,16,1,0,0.6103800013661385
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,64,1,0,0.5649439990520477
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,32,1,0,0.5999119877815247
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,128,1,0,0.5605559945106506
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,64,1,0,0.5919920057058334
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,8192,1,0,1.808440014719963
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,256,1,0,0.5647200047969818
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,128,1,0,0.6152920052409172
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,512,1,0,0.577380008995533
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,256,1,0,0.6529360041022301
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,16384,1,0,5.14190000295639
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,1024,1,0,0.6334559917449951
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,4096,1,0,4.8069000244140625
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,512,1,0,0.7338400185108185
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,4,1,0,0.5937279909849167
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,2048,1,0,0.7893359959125519
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,8,1,0,0.5865359976887703
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,1024,1,0,1.1056479960680008
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,4096,1,0,1.3288920223712921
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,16,1,0,0.5839520022273064
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,32,1,0,0.5871360003948212
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,16384,1,0,5.949940025806427
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,64,1,0,0.5885320007801056
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,2048,1,0,2.213220030069351
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,128,1,0,0.5859159976243973
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,4,1,0,0.5637639984488487
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,256,1,0,0.5888519957661629
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,8,1,0,0.5717040002346039
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,8192,1,0,3.2754360139369965
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,512,1,0,0.6131520122289658
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,16,1,0,0.5708080008625984
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,1024,1,0,0.6504600048065186
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,32,1,0,0.5692120119929314
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,64,1,0,0.5704879909753799
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,2048,1,0,0.8227559998631477
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,128,1,0,0.5726680085062981
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,256,1,0,0.5771880000829697
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,4096,1,0,1.4215880036354065
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,4096,1,0,5.148880064487457
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,512,1,0,0.5968439877033234
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,4,1,0,0.6208359971642494
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,1024,1,0,0.6387199908494949
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,8,1,0,0.6221599951386452
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,2048,1,0,0.7841519936919212
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,16,1,0,0.6211439892649651
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,8192,1,0,3.478987991809845
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,4096,1,0,1.3641639947891235
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,32,1,0,0.6257119998335838
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,8192,1,0,13.74594795703888
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,64,1,0,0.643499992787838
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,128,1,0,0.6951520070433617
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,16384,1,0,10.528948068618774
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,256,1,0,0.7674079909920692
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,4,1,0,0.5607439875602722
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,8,1,0,0.5710480064153671
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,8192,1,0,5.027176052331924
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,512,1,0,1.1067039966583252
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,16,1,0,0.5684560015797615
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,32,1,0,0.5864880084991455
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,64,1,0,0.5635359957814217
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,128,1,0,0.5659880116581917
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,1024,1,0,2.0352440178394318
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,256,1,0,0.5923440009355545
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,512,1,0,0.6302760019898415
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,1024,1,0,0.741315983235836
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,16384,1,0,10.82152795791626
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,2048,1,0,1.1080520153045654
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,4,1,0,0.5940359979867935
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,8,1,0,0.6004079952836037
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,2048,1,0,4.280492007732391
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,16,1,0,0.5901280045509338
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,32,1,0,0.5935920029878616
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,4096,1,0,2.4273360073566437
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,64,1,0,0.5934359952807426
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,128,1,0,0.5962639972567558
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,256,1,0,0.6246640011668205
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,512,1,0,0.6561920046806335
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,1024,1,0,0.7670399993658066
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,2048,1,0,1.2060199826955795
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,16384,1,0,40.94174003601074
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,8192,1,0,6.606060087680817
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,4,1,0,0.5919400155544281
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,16384,1,0,18.134963989257812
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,4096,1,0,2.6497199833393097
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,8,1,0,0.5928399935364723
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,4,1,0,0.5705279931426048
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,16,1,0,0.5835400000214577
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,8,1,0,0.573172003030777
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,4096,1,0,10.597324132919312
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,32,1,0,0.5955440029501915
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,16,1,0,0.5743239969015121
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,64,1,0,0.6292559877038002
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,32,1,0,0.5682080015540123
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,4,1,0,0.6874520033597946
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,128,1,0,0.66006800532341
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,64,1,0,0.5689719989895821
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,256,1,0,0.73973198980093
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,8,1,0,0.6853160038590431
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,128,1,0,0.5717559978365898
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,16,1,0,0.6915439963340759
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,512,1,0,0.975088007748127
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,256,1,0,0.5951960012316704
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,32,1,0,0.7090759947896004
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,512,1,0,0.6446159929037094
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,1024,1,0,0.7311359867453575
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,64,1,0,0.7400639951229095
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,1024,1,0,1.7858880013227463
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,8192,1,0,7.18147999048233
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,128,1,0,0.8266519978642464
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,2048,1,0,1.6008480042219162
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,256,1,0,1.130623996257782
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,2048,1,0,3.748988002538681
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,512,1,0,1.9824040085077286
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,4096,1,0,4.702592074871063
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,1024,1,0,3.907499998807907
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,16384,1,0,20.32359218597412
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,16384,1,0,41.426199436187744
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,4,1,0,0.5817799940705299
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,4,1,0,0.5937319993972778
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,8,1,0,0.5783880054950714
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,4096,1,0,9.633739948272705
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,8,1,0,0.6046280041337013
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,16,1,0,0.5778119936585426
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,16,1,0,0.5963839888572693
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,32,1,0,0.5546000003814697
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,32,1,0,0.6071040108799934
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,64,1,0,0.5707439929246902
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,64,1,0,0.6279880031943321
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,128,1,0,0.5920000076293945
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,128,1,0,0.6787400022149086
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,256,1,0,0.6247800067067146
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,256,1,0,0.740619994699955
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,512,1,0,0.7159719839692116
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,8192,1,0,11.62852793931961
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,2048,1,0,8.676715850830078
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,512,1,0,1.0714159905910492
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,1024,1,0,1.007123976945877
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,4,1,0,0.7907159999012947
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,1024,1,0,1.8515560030937195
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,2048,1,0,1.9804079979658127
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,8,1,0,0.7911479994654655
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,16384,1,0,21.658655881881714
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,16,1,0,0.8184319958090782
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,4,1,0,0.6156880110502243
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,32,1,0,0.8611760064959526
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,8,1,0,0.6060599908232689
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,16,1,0,0.6096480041742325
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,2048,1,0,3.8976680040359497
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,64,1,0,0.9444959983229637
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,32,1,0,0.6114920005202293
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,4096,1,0,4.661523938179016
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,64,1,0,0.6113880053162575
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,128,1,0,1.2397120147943497
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,128,1,0,0.6320880055427551
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,256,1,0,0.6714479997754097
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,512,1,0,0.7558759972453117
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,256,1,0,2.0935560166835785
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,1024,1,0,1.1071560084819794
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,2048,1,0,2.2197199761867523
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,512,1,0,3.84306800365448
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,4096,1,0,9.92662787437439
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,4,1,0,0.6330559998750687
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,8,1,0,0.6338840052485466
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,16,1,0,0.6397400051355362
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,4096,1,0,5.139679968357086
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,32,1,0,0.6617040038108826
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,8192,1,0,26.13287615776062
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,8192,1,0,13.297775864601135
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,4,1,0,0.6438039988279343
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,64,1,0,0.710391990840435
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,128,1,0,0.7595480009913445
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,8,1,0,0.644167996942997
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,16,1,0,0.6387679949402809
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,256,1,0,1.0269279778003693
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,32,1,0,0.6391960009932518
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,1024,1,0,7.799475967884064
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,64,1,0,0.6529560014605522
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,512,1,0,1.7749439775943756
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,128,1,0,0.6914560049772263
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,4,1,0,1.0295440107584
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,256,1,0,0.7930839955806732
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,512,1,0,1.098696008324623
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,8,1,0,1.0513519942760468
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,1024,1,0,3.4598160088062286
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,16,1,0,1.0996760129928589
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,1024,1,0,2.041491985321045
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,32,1,0,1.1907120198011398
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,64,1,0,1.4614159911870956
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,2048,1,0,4.308696031570435
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,128,1,0,2.2798800468444824
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,2048,1,0,7.8140639662742615
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,256,1,0,4.026308000087738
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,4096,1,0,10.625216007232666
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,4,1,0,0.6908599957823753
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,16384,1,0,53.67154407501221
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,8,1,0,0.6897759959101677
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,512,1,0,7.5926119685173035
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,4,1,0,0.5792759954929352
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,16,1,0,0.6926399990916252
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,8,1,0,0.5842920020222664
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,32,1,0,0.7048439979553223
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,16,1,0,0.5786919891834259
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,64,1,0,0.7479480057954788
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,32,1,0,0.5875959917902946
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,4,1,0,1.5659840106964111
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,128,1,0,0.816611997783184
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,64,1,0,0.5838800072669983
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,128,1,0,0.6002360060811043
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,256,1,0,1.1281680017709732
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,256,1,0,0.6501680016517639
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,8,1,0,1.6056159883737564
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,512,1,0,0.7079480066895485
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,512,1,0,1.96942800283432
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,1024,1,0,1.09128800034523
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,4096,1,0,19.430376291275024
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,16,1,0,1.701571986079216
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,4,1,0,0.6998240053653717
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,8,1,0,0.6946600079536438
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,2048,1,0,2.8368280231952667
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,32,1,0,1.9339919835329056
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,16,1,0,0.7118800133466721
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,16384,1,0,40.663424491882324
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,1024,1,0,3.8714919984340668
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,32,1,0,0.7587919980287552
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,4,1,0,0.592743992805481
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,64,1,0,0.8330239951610565
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,8,1,0,0.5703920051455498
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,64,1,0,2.714212030172348
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,128,1,0,1.109607994556427
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,16,1,0,0.5815080031752586
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,32,1,0,0.5822639912366867
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,64,1,0,0.6066000014543533
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,256,1,0,1.8183520138263702
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,128,1,0,0.651311993598938
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,256,1,0,0.7369559928774834
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,4096,1,0,7.958888053894043
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,512,1,0,0.9667279943823814
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,128,1,0,4.376999914646149
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,512,1,0,3.3212599754333496
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,1024,1,0,1.7667120099067688
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,2048,1,0,8.743039846420288
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,4,1,0,0.7972840070724487
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,8,1,0,0.8008200004696846
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,2048,1,0,3.7475920021533966
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,16,1,0,0.8242960050702095
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,32,1,0,0.8598399981856346
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,1024,1,0,6.9903000593185425
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,64,1,0,0.9602599963545799
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,256,1,0,7.901095986366272
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,128,1,0,1.2325200140476227
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,256,1,0,2.0788040161132812
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,4,1,0,2.7056719958782196
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,4096,1,0,9.53764796257019
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,512,1,0,3.8092239797115326
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,8,1,0,2.84128001332283
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,8192,1,0,24.55748414993286
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,16,1,0,2.9955440163612366
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,2048,1,0,15.90739619731903
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,1024,1,0,7.732172012329102
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,4,1,0,0.881915993988514
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,8,1,0,0.8606599941849709
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,4,1,0,1.0718520134687424
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,32,1,0,3.7286760210990906
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,16,1,0,0.9251199960708618
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,8,1,0,1.0834240019321442
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,32,1,0,0.983135998249054
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,16,1,0,1.1301239877939224
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,64,1,0,1.2177680134773254
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,32,1,0,1.218256026506424
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,16384,1,0,82.99623203277588
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,128,1,0,1.9087039977312088
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,64,1,0,1.4755719900131226
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,64,1,0,5.312063992023468
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,4,1,0,0.6082759946584702
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,8,1,0,0.6119159981608391
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,16,1,0,0.6060959994792938
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,128,1,0,2.2693880200386047
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,32,1,0,0.6384159997105598
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,256,1,0,3.3770560026168823
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,64,1,0,0.6745679974555969
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,128,1,0,0.7603319957852364
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,256,1,0,0.9798519909381866
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,8192,1,0,26.248923778533936
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,256,1,0,3.9803119897842407
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,512,1,0,1.7035239934921265
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,512,1,0,6.561631917953491
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,128,1,0,8.747739791870117
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,1024,1,0,3.3065600097179413
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,512,1,0,7.449940025806427
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,2048,1,0,7.536823987960815
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,4,1,0,1.6297599971294403
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,8,1,0,1.6568560004234314
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,16,1,0,1.7626359909772873
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,1024,1,0,14.07099199295044
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,32,1,0,1.9721879810094833
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,4,1,0,1.2104920148849487
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,8,1,0,1.266427993774414
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,64,1,0,2.7137359976768494
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,16,1,0,1.3442880064249039
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,32,1,0,1.5430039912462234
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,128,1,0,4.334692001342773
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,64,1,0,2.1908759772777557
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,4096,1,0,18.76860761642456
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,4,1,0,0.6346039921045303
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,128,1,0,3.5888999700546265
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,8,1,0,0.6402800008654594
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,16,1,0,0.6599999964237213
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,32,1,0,0.6969840079545975
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,64,1,0,0.7868679985404015
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,128,1,0,1.0002119913697243
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,256,1,0,7.764703989028931
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,256,1,0,1.7026959955692291
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,256,1,0,6.624819993972778
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,512,1,0,3.1234599947929382
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,4,1,0,2.7777840197086334
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,8,1,0,2.8612719774246216
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,1024,1,0,6.723823964595795
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,16,1,0,3.1073480248451233
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,512,1,0,13.25273609161377
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,32,1,0,3.8096239268779755
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,4,1,0,1.921312004327774
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,8,1,0,2.017552003264427
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,64,1,0,5.28874796628952
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,16,1,0,2.1916719675064087
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,2048,1,0,15.271956086158752
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,32,1,0,2.78643199801445
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,64,1,0,4.099976062774658
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,128,1,0,8.541751980781555
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,16384,1,0,105.56857967376709
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,128,1,0,7.042371988296509
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,4,1,0,0.6003520116209984
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,8,1,0,0.5923280045390129
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,16,1,0,0.5959480106830597
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,32,1,0,0.6001719981431961
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,64,1,0,0.622859999537468
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,16384,1,0,83.32479190826416
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,128,1,0,0.6605079993605614
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,256,1,0,0.815200001001358
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,4,1,0,0.6056519970297813
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,8,1,0,0.6016039997339249
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,512,1,0,1.047444000840187
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,16,1,0,0.61234400421381
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,32,1,0,0.6236959919333458
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,64,1,0,0.6732039898633957
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,1024,1,0,2.9961239993572235
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,128,1,0,0.7564959973096848
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,256,1,0,0.9748800024390221
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,512,1,0,1.7126080095767975
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,256,1,0,13.34441602230072
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,2048,1,0,6.331112027168274
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,1024,1,0,3.2761479914188385
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,4,1,0,3.3145639896392822
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,4096,1,0,37.913700103759766
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,2048,1,0,7.506331980228424
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,8,1,0,3.521791994571686
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,4,1,0,0.7617879956960678
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,8,1,0,0.7641600072383881
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,16,1,0,0.8263280093669891
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,32,1,0,0.9642879962921143
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,16,1,0,4.265448033809662
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,64,1,0,1.0879400223493576
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,128,1,0,1.76494000852108
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,4096,1,0,15.99831998348236
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,4,1,0,0.6136839911341667
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,8,1,0,0.6375119909644127
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,256,1,0,3.143316000699997
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,16,1,0,0.625795990228653
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,32,1,0,5.4477479457855225
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,32,1,0,0.6508840024471283
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,64,1,0,0.693571999669075
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,128,1,0,0.8364720046520233
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,256,1,0,1.3626919835805893
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,512,1,0,6.183491945266724
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,4096,1,0,18.902692317962646
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,512,1,0,3.1635559797286987
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,4,1,0,0.6601440086960793
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,64,1,0,8.144052147865295
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,8,1,0,0.6746639981865883
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,16,1,0,0.687948003411293
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,32,1,0,0.7364999949932098
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,64,1,0,0.8126360028982162
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,1024,1,0,5.323028028011322
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,128,1,0,1.0372599884867668
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,256,1,0,1.75347600877285
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,1024,1,0,13.451528072357178
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,128,1,0,14.04556393623352
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,2048,1,0,15.22518801689148
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,512,1,0,3.207311987876892
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,1024,1,0,6.589215993881226
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,2048,1,0,15.280187964439392
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,2048,1,0,30.29206418991089
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,4,1,0,1.0036799982190132
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,8,1,0,1.038992002606392
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,16,1,0,1.146816000342369
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,32,1,0,1.3524280041456223
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,64,1,0,1.9555520117282867
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,128,1,0,3.251316010951996
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,4096,1,0,51.20275592803955
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,4,1,0,0.7519480064511299
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,8,1,0,0.6911960020661354
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,16,1,0,0.7128679975867271
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,256,1,0,6.215640068054199
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,32,1,0,0.7596680000424385
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,64,1,0,0.8349559977650642
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,128,1,0,1.1424079835414886
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,256,1,0,2.599923998117447
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,4096,1,0,38.00896406173706
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,4,1,0,0.7518960013985634
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,512,1,0,5.917800009250641
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,8,1,0,0.7892560064792633
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,16,1,0,0.8339480087161064
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,512,1,0,12.531275868415833
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,32,1,0,0.8996880054473877
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,64,1,0,1.1048559993505478
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,128,1,0,1.7943799942731857
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,256,1,0,3.250632017850876
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,1024,1,0,11.31687593460083
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,512,1,0,6.351671993732452
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,1024,1,0,26.659523963928223
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,1024,1,0,13.40066385269165
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,4,1,0,1.621476024389267
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,8,1,0,1.5665999799966812
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,16,1,0,1.871156007051468
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,2048,1,0,27.347327947616577
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,32,1,0,2.3573239743709564
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,4,1,0,0.9226559996604919
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,8,1,0,0.894899994134903
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,16,1,0,0.9268999993801117
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,64,1,0,3.600928008556366
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,32,1,0,0.9920400157570839
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,64,1,0,1.3129199743270874
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,128,1,0,2.3742600083351135
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,128,1,0,6.302563965320587
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,256,1,0,4.634952068328857
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,2048,1,0,30.463307857513428
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,256,1,0,12.331907987594604
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,512,1,0,12.067035794258118
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,4,1,0,0.9897559881210327
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,8,1,0,1.0289799869060516
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,16,1,0,1.1531360000371933
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,32,1,0,1.3177000135183334
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,64,1,0,1.9848680198192596
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,128,1,0,3.316980004310608
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,256,1,0,6.306432008743286
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,512,1,0,24.967624187469482
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,4,1,0,2.5028719902038574
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,512,1,0,12.759552121162415
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,8,1,0,2.6326359808444977
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,1024,1,0,36.03471231460571
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,4,1,0,1.2954120337963104
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,16,1,0,3.231880009174347
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,8,1,0,1.304616004228592
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,16,1,0,1.3875439763069153
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,32,1,0,4.440095961093903
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,32,1,0,1.6674160063266754
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,64,1,0,3.199176013469696
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,64,1,0,7.0951679944992065
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,128,1,0,7.481799960136414
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,1024,1,0,27.3338680267334
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,128,1,0,12.723708033561707
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,4,1,0,1.5021920055150986
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,256,1,0,14.974712014198303
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,8,1,0,1.5629760175943375
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,16,1,0,1.9203039854764938
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,32,1,0,2.32397198677063
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,64,1,0,3.58433997631073
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,128,1,0,6.32915598154068
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,256,1,0,24.63225221633911
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,512,1,0,30.297736167907715
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,4,1,0,4.43666809797287
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,4,1,0,2.0746119916439056
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,256,1,0,12.47350800037384
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,8,1,0,2.18792000412941
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,8,1,0,5.090655982494354
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,16,1,0,2.392787992954254
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,16,1,0,6.2338080406188965
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,32,1,0,3.697620004415512
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,64,1,0,7.1563640832901
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,32,1,0,8.624499917030334
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,512,1,0,26.04788827896118
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,128,1,0,13.745172023773193
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,4,1,0,2.2827799916267395
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,64,1,0,14.003727912902832
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,8,1,0,3.008839935064316
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,16,1,0,3.0955840349197388
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,32,1,0,4.291139900684357
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,64,1,0,6.882740020751953
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,256,1,0,27.231339693069458
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,128,1,0,25.470579862594604
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,4,1,0,3.5079639852046967
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,128,1,0,12.602952003479004
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,8,1,0,3.7401880025863647
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,16,1,0,5.042400002479553
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,32,1,0,7.5836599469184875
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,256,1,0,25.81828784942627
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,4,1,0,4.118988037109375
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,64,1,0,11.409972071647644
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,8,1,0,4.714192032814026
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,16,1,0,5.941212058067322
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,32,1,0,8.40694785118103
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,128,1,0,22.77294397354126
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,64,1,0,13.6421560049057
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,128,1,0,25.276779651641846
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,1,1,0,0.558396004140377
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,4,1,0,0.5595039948821068
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,8,1,0,0.5536119937896729
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,16,1,0,0.5545400083065033
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,32,1,0,0.6092480048537254
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,64,1,0,0.5538880005478859
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,128,1,0,0.5608840063214302
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,256,1,0,0.5538440048694611
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,512,1,0,0.5966960042715073
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,1024,1,0,0.7217440009117126
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,2048,1,0,1.148276001214981
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,4096,1,0,2.3834240436553955
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,8192,1,0,6.553268015384674
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,1,1,0,0.5653679892420769
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,16384,1,0,20.240612030029297
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,4,1,0,0.5587359964847565
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,4,1,0,0.5701679959893227
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,8,1,0,0.5494400039315224
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,8,1,0,0.5727439969778061
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,16,1,0,0.5445879995822906
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,16,1,0,0.569531999528408
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,32,1,0,0.5517480000853539
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,32,1,0,0.5661080032587051
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,64,1,0,0.5470200031995773
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,64,1,0,0.5749599933624268
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,128,1,0,0.5522719994187355
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,128,1,0,0.5729199945926666
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,256,1,0,0.5589319989085197
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,256,1,0,0.6104800030589104
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,512,1,0,0.5965880081057549
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,512,1,0,0.6948360055685043
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,1024,1,0,0.7084000036120415
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,1024,1,0,1.0435639917850494
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,2048,1,0,1.1378959864377975
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,2048,1,0,1.9325080066919327
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,4096,1,0,2.3402600288391113
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,4096,1,0,4.757652044296265
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,8192,1,0,6.40394002199173
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,8192,1,0,13.186716079711914
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,16384,1,0,20.068419933319092
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,4,1,0,0.55240398645401
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,8,1,0,0.5558480024337769
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,16,1,0,0.5618919879198074
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,32,1,0,0.5587479993700981
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,64,1,0,0.5772720128297806
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,128,1,0,0.5939399972558022
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,256,1,0,0.6130800023674965
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,512,1,0,0.6950560063123703
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,1024,1,0,1.0340960025787354
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,2048,1,0,1.9478799849748611
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,4096,1,0,4.595480024814606
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,8192,1,0,13.289344072341919
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,16384,1,0,40.73720455169678
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,4,1,0,0.5823760032653809
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,8,1,0,0.5762360021471977
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,16,1,0,0.5721400007605553
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,32,1,0,0.5745679959654808
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,64,1,0,0.581836000084877
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,128,1,0,0.6253839954733849
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,256,1,0,0.7066959962248802
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,512,1,0,1.018364004790783
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,1024,1,0,1.7262880057096481
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,2048,1,0,3.773236006498337
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,4096,1,0,9.566491961479187
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,16384,1,0,39.631508350372314
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,4,1,0,0.5613840073347092
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,8,1,0,0.5607640072703362
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,16,1,0,0.5679919943213463
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,8192,1,0,26.23850417137146
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,32,1,0,0.5565440058708191
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,64,1,0,0.5605279952287674
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,128,1,0,0.6039640009403229
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,256,1,0,0.6894200146198273
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,512,1,0,1.001595988869667
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,1024,1,0,1.7211679965257645
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,2048,1,0,3.6756439805030823
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,4096,1,0,9.187932133674622
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,8192,1,0,25.654240131378174
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,16384,1,0,83.5085678100586
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,4,1,0,0.7457800060510635
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,8,1,0,0.9090479984879494
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,16,1,0,0.7262439951300621
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,32,1,0,0.7236240059137344
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,64,1,0,0.7530919909477234
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,128,1,0,0.837332010269165
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,256,1,0,1.0848159790039062
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,512,1,0,1.643992006778717
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,1024,1,0,3.254759967327118
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,2048,1,0,7.7286399602890015
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,4096,1,0,18.426432132720947
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,16384,1,0,82.2240800857544
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,4,1,0,0.5698199942708015
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,8,1,0,0.5671879947185516
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,16,1,0,0.5689999982714653
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,32,1,0,0.5761400014162064
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,64,1,0,0.6341480016708374
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,128,1,0,0.71373600512743
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,256,1,0,0.9960720017552376
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,512,1,0,1.6484559923410416
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,1024,1,0,3.222691982984543
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,2048,1,0,7.285516083240509
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,4096,1,0,18.386451959609985
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,8192,1,0,51.885796546936035
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,4,1,0,0.6160280108451843
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,8,1,0,0.6200999915599823
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,16,1,0,0.6336839944124222
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,32,1,0,0.6479919999837875
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,64,1,0,0.739827997982502
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,128,1,0,1.06216399371624
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,256,1,0,1.7035119980573654
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,512,1,0,3.113396018743515
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,1024,1,0,6.342676043510437
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,8192,1,0,54.26512384414673
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,2048,1,0,15.075147867202759
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,4096,1,0,38.07896375656128
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,4,1,0,0.709587998688221
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,8,1,0,0.7194439917802811
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,16,1,0,0.7757959887385368
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,32,1,0,0.9034359902143478
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,64,1,0,1.09859998524189
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,128,1,0,1.6532279998064041
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,256,1,0,3.0476480424404144
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,512,1,0,6.1549999713897705
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,1024,1,0,13.369759917259216
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,2048,1,0,30.051495790481567
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,4,1,0,0.8714479953050613
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,8,1,0,0.9073280021548271
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,16,1,0,0.9883359968662262
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,32,1,0,1.2398599982261658
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,64,1,0,1.7949640154838562
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,128,1,0,3.11940798163414
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,256,1,0,5.893996059894562
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,512,1,0,12.306944131851196
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,16384,1,0,167.35549545288086
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,4,1,0,0.6018919944763184
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,8,1,0,0.6030400097370148
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,16,1,0,0.6063520088791847
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,32,1,0,0.6443119943141937
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,64,1,0,0.7290639877319336
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,128,1,0,1.0204760134220123
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,1024,1,0,26.359219789505005
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,256,1,0,1.6651960015296936
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,512,1,0,3.115400016307831
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,4,1,0,1.240383967757225
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,8,1,0,1.333755999803543
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,1024,1,0,6.5018839836120605
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,16,1,0,1.573752000927925
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,32,1,0,2.1037079989910126
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,64,1,0,3.368996024131775
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,128,1,0,6.054976046085358
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,2048,1,0,14.864260077476501
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,256,1,0,12.029128193855286
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,512,1,0,24.517595767974854
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,4096,1,0,36.80549192428589
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,4,1,0,0.6632079929113388
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,8,1,0,0.6516079977154732
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,16,1,0,0.7004239931702614
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,4,1,0,1.9939080029726028
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,32,1,0,0.8204199895262718
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,64,1,0,1.0691480040550232
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,8,1,0,2.2525080144405365
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,128,1,0,1.678871989250183
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,16,1,0,2.768467962741852
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,256,1,0,3.0940680503845215
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,32,1,0,3.9726999700069427
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,512,1,0,6.071164011955261
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,64,1,0,6.625156044960022
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,1024,1,0,13.03961193561554
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,128,1,0,11.997356176376343
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,256,1,0,23.652899980545044
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,2048,1,0,29.301648139953613
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,4,1,0,3.5885920226573944
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,8,1,0,4.095567882061005
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,16,1,0,5.282931983470917
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,32,1,0,7.905784070491791
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,64,1,0,13.137744188308716
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,128,1,0,24.075455904006958
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,4096,1,0,77.72455215454102
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1024,4,1,0,6.726135969161987
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,4,1,0,0.8844959959387779
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,8,1,0,0.8835120052099228
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,16,1,0,0.9609360024333
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,32,1,0,1.1986920088529587
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,64,1,0,1.7943000048398972
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1024,8,1,0,7.988495945930481
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,128,1,0,3.1297399699687958
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,256,1,0,6.13643205165863
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1024,16,1,0,10.381420016288757
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,512,1,0,12.220395922660828
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1024,32,1,0,15.445095896720886
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,1024,1,0,25.769299745559692
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1024,64,1,0,26.166491746902466
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,2048,1,0,59.95259189605713
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,4,1,0,1.1771080046892166
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,8,1,0,1.2554839998483658
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,16,1,0,1.4947080165147781
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,32,1,0,2.039192020893097
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,64,1,0,3.320939987897873
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,128,1,0,6.11574000120163
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,256,1,0,12.059915900230408
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,512,1,0,25.066275596618652
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,1024,1,0,54.49678087234497
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,4,1,0,1.8888639956712723
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,8,1,0,2.0915440022945404
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,16,1,0,2.65406796336174
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,32,1,0,3.8756199777126312
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,64,1,0,6.580884039402008
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,128,1,0,11.983004093170166
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,256,1,0,24.331599950790405
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,512,1,0,52.0608606338501
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,4,1,0,3.310824006795883
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,8,1,0,3.842527985572815
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,16,1,0,5.058791995048523
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,32,1,0,7.402884006500244
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,64,1,0,12.770524024963379
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,128,1,0,23.874127864837646
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,256,1,0,49.653432846069336
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,4,1,0,6.2547120451927185
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,8,1,0,7.575136065483093
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,16,1,0,9.871851921081543
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,32,1,0,14.83409583568573
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,64,1,0,25.39435577392578
SGLang,0.5.8.post1,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,128,1,0,49.28502798080444
